Full text of "Extraterrestrial Civilizations Problems Of Interstellar Communication"

See other formats

Problems of
Interstellar
Communication

S. A. Kaplan, Editor

EXTRATERRESTRIAL CIVILIZATION

Problems of Interstellar Communication

Translated from Russian

Published for the National Aeronautics and Space Administration
and the National Science Foundation, Washington, D.C.
by the Israel Program for Scientific Translations

S. A. KAPLAN, Editor

EXTRATERRESTRIAL CIVILIZATIONS

Problems of Interstellar Communication

(Vnezemnye tsivilizatsii.
Problemy mezhzvezdnoi svyazi)

Izdatel' stvo "Nauka"
Glavnaya Redaktsiya
Fiziko-Matematicheskoi Literatury
Moskva 1969

Translated from Russian

Israel Program for Scientific Translations
Jerusalem 1971

TT 70- 50081
NASA TT F-631

Published Pursuant to an Agreement with
THE NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
and
THE NATIONAL SCIENCE FOUNDATION, WASHINGTON, D. C.

Translated by IPST staff

Printed in Jerusalem by Keter Press
Binding: Wiener Bindery Ltd., Jerusalem

Available from the
U.S. DEPARTMENT OF COMMERCE
National Technical Information Service
Springfield, Va. 22151

X/16/3

Table of Contents

Introduction; EXOSOCIOLOGY — THE SEARCH FOR SIGNALS FROM
EXTRATERRESTRIAL CIVILIZATIONS (S. A. Kaplan)

The theory of development of civilizations (3). The search

for signals from extraterrestrial civilizations (5), De-

coding aspects of the program of search for extraterrestrial

civilizations (8).

Bibliography (ees ee esee sonehhocoesoces tton

Chapter I: THE ASTROPHYSICAL ASPECT OF THE SEARCH FOR
SIGN ALS FROM EXTRATERRESTRIAL CIVILIZA TIONS

(N.S.Kardashev) ... cece ccc creer nerve tmn
$1. Introduction ....ccccccs sneer tcc ccs E S
$2. The Main Dilemma ......cccscecccvscvcccscvece
$3. The Completeness and Reliability of Modern Astrophysical Data
$4, Civilizations and the Main Features of their Development ....

$5. The Search for Signs of Activity of Supercivilizations ..
Energy sources (28), Solid matter (39).

$6. The Search for Information Transmissions ..........

$7. The Program of Search for Supercivilizations ........

Bibliography .... ee eee heeheehehhhr hors

12
12
14
15
22
28

42
55
57

Chapter II: THE EFFECT OF THE SPACE MEDIUM ON THE PROPAGATION
OF RADIO SIGNALS (B. N,Panovkin .. cece seer cere eee
Bibliography ..ccccsecesescerernsesssessvavsesereves

Chapter III: THE POSSIBILITY OF RADIO COMMUNICATION WITH

EXTRATERRESTRIAL CIVILIZATIONS (L.M.Gindilis) ......

$1, Elements of the General Theory of Communication .......06.
Structure and fundamental characteristics of a communication
system (68). Quantitative definition of information (70).
Transformation of a message into a signal. Forms of modulation
(72), Physical characteristics of signals (73), Relation of
pulse length to pulse band width. Number of pulses trans-
mitted through a channel of given band width A f(77).
Transmission of continuous functions by pulsed signals (78).
Transinission rate of a communication channel (82).

iii

$2.

$3.
§ 4,

Range and Information Content of Interstellar Communication .....
The optimum communication frequencies (86). Range of
communication (88). Range of detection (95). Range of
reception of pulse signals (99), Length of transmission,
Directivity and information content (100),

Call Signals and Artificiality Criteria ,...... ecce eee

Methods of Detection of EC Signals ......-cccvcccvcrcccee
Transmitter power. The power potential of a civilization (109).
Radio communication between galaxies (118), Monochromatic
signals, Frequency scanning (120), Direction scanning (125).
Wide-band signals. Sky surveys (127).

Bibliography serros redie esee eee eee ttn

Chapter IV. METHODS OF MESSAGE DECODING (B. V. Sukhotin) .......

$1.
$2.

$3.

$4.

$5.

$6.

$7.

$8.

$9.

Introduction . sce ceec ssc cr en rereseseescescvvesens

The Concept of a Message, Its Intelligibility and Meaningfulness ..
Definition of message (135). Artificial and natural messages
(136). Intelligibility of a message (137). Meaningfulness
of a message, predictive system, language (139).

Traditional Methods of Military and Linguistic Deciphering ......
Military deciphering (140), Linguistic deciphering (143).

Sequence of Application and Structure of Decoding Algorithms ....
Sequence of algorithm application. Levels (144).

Structure of algorithms: sets of alternatives, quality function,

computation procedures, Types of algorithms (148).
Classification Algorithms (Part D ,, ecce eee hh

Distinctive features and classifications (151). Algorithms

for the "identification of vowels and consonants (152),

Matching Algorithms (Part D. ...cerccccceveceseceveces
Algorithms identifying code sequences (160), An example
illustrating the application of the concept of meaningfulness (164),

Pattern Decoding Algorithms s. eee eee ee eee eet nn
Language of images, Connectedness and detailedness (160).

The language of quality functions. Some procedures (169).
Algorithms Analogous to Algorithms which Construct Bilingual
Dictionaries, , e. eese eese eeehh shoe

Letter -comparison algorithms using the properties of close

neighborhoods (176). An algorithm using distant neighborhoods

(188).

Classification Algorithms (End), e. eee eee
"Mathematically" correct algorithm for vowel-and-consonant
identification (185), An algorithm translating syllabic writing
into alphabet writing (188), An algorithm for "semantic"
classification of words (192).

103
109

140

144

151

160

166

176

185

$10, Closeness-Identifying Algorithms ...... ce ceeeee rrt

Algorithm determining the graph of syntactic connections

of words in a sentence (195). An algorithm identifying
"types of syntactic relationship" of words (198). The
simplest algorithm of literal machine translation (200).

$11, Matching Algorithms (End ........... eer

$12. Conclusion

Morpheme -identifying algorithms (202), Letter-identifying

algorithms (207).

Ce cssoosoooooosoo’‘o’ooooooo’on‘n’o‘oo’‘o‘’‘’l I’

Bibliography . e.ssesesossosooosessoosoooososooosoo

Chapter V. RATES OF DEVELOPMENT OF CIVILIZATIONS AND THEIR

FORECASTING (G. M. Khovanov) ,.. c.c celere
The Importance of the Problem of Rates of Development ....
The Aspects of Development of Civilizations .......... e

$1.
$2.

$3.

$4.
$5.

Language and communication (215), Demographic

characteristics of civilization (217), The development of

individual abilities (218).

Indices of Technical Progress ...... eee enn
On the succession of indices (221). Mathematical functions

describing growth rates (223).

Rates of Growth of Science... ellen

Forecasting .. cece ccc cee r ersten teens etnscesesetes
Classification of forecasts (229). Accuracy of forecasts (230).
Forecasting the rates of scientific and technological progress
(231). Forecasting the growth rates of the Earth civilization

(233).

Bibliography ., cece cere re ccce se vecerererssccves

Chapter VI, SOME GENERAL TOPICS OF THE PROBLEM OF EXTRA-

$1.
$2.

$3,

$4.

TERRESTRIAL CIVILIZATIONS ....... e een
Introduction ,.. eee eee at
The Methodology of the "Radio Astronomical" Aspect of the
Problem. The "Energy" Hypothesis ,. 4... eee
An Alternative Point of View, S, Lem and His Summa
Technologiae ... «ee eeeeeeeeeeee sehen

The Problem of Extraterrestrial Civilizations from the Point of

View of the General Theory of Systems ... eee

Bibitography €*890208400692.8208069006889529422028952808025096028492820690€908

185

202

211
212

213

213
214

220

224
228

236

238

239

247

253
264

Introduction

EXOSOCIOLOGY — THE SEARCH FOR SIGNALS
FROM EXTRATERRESTRIAL CIVILIZATIONS

The search for signals from extraterrestrial civilizations is one of the
most intriguing problems raised by modern science. What are these signals,
where and how are we to look for them, and should we devote time and effort
to this search? These questions were originally in the domain of science
fiction, and it is only recently that they began to be considered seriously by
astronomers, physicists, biologists, linguists, and philosophers in scientific
conferences and in various articles and books. However, despite the
numerous questions raised and the various hypotheses advanced, there has
been very little real scientific research in this direction. Even the cardinal
question of the actual outcome of the encounter of mankind with extraterres-
trial civilization — whether it will be beneficial or harmful — has not been
answered unanimously. It suffices to mention how the excessively optimistic
prospects of interstellar communication drawn by I. A. Efremov in his
"Andromeda Nebula" contrast with the distressing picture envisaged by
F. Hoyle and Ch. Elliott in their "A for Andromeda." Incidentally, both these
books were written by eminent scientists, fully conversant with the grave
implications of the problem, and not by professional science-fiction writers
who sometimes stand accused of flippant treatment of the subject.

The expansion of man into outer space led to a rapid development of
new branches of science and technology. One of these new disciplines is
exobiology, a science dealing with the origin and evolution of life under
extraterrestrial conditions. The very wide range of topics considered by
exobiology attracted the attention of scientists from a variety of fields. Some
problems of exobiology are even now nearing final solution, whereas others
are still in the embryonic stages of research.

One of the fundamental problems of exobiology is purely astronomical:
what is the probability of any individual star being surrounded by planets
with life-sustaining conditions? In other words, the primary task is to find
the probability of existence of a planet with a mass not radically different
from the Earth mass, adequate axial rotation parameters, and an atmo-
sphere, which lies in the "life-sustaining heat zone," i.e., not too far from
the primary to be permanertly frozen and yet not too near it for the surface
to be scorched. Although there is a measure of uncertainty in the very
formulation of the problem, a more or less definite solution has been
obtained by now. The probability of the existence of such a life-sustaining
planet is of the order of a few percent.

However, the existence of conditions which are potentially capable of
sustaining life does not imply that life actually exists on the particular planet.

EXTRA TERRESTRIAL CIVILIZATIONS

Unfortunately, there is a great deal of uncertainty in this purely biological
aspect of the matter. Most authorities seem to be of the opinion that the
probability of life inception is fairly high. There is, however, an alternative
point of view, equally valid in terms of the actual proof available (or rather
totallack thereof), which suggests that the probability of life inception is
negligible even under ideally suitable conditions. The extreme difficulty of
the problem is further aggravated by the lack of a reliable theory of the
origin of life on Earth. The attempts to reproduce this process in laboratory
have failed so far. The discovery of even minute traces of life on Venus or
Mars or the demonstration of repetitiveness or multiplicity of the process of
life inception on Earth would provide invaluable information toward the
solution of the problem (together with laboratory research). Therefore we
do not exaggerate if we say that the success of exobiology in the solution of
its fundamental problem — elucidating the possibility of life originating under
certain conditions — largely depends on the level of our space technology.
The rapid advances of modern astronautics instill us with hope that the
probability of life on a planet endowed with appropriate conditions will be
determined in the nearest future.

No less complex and equally far from solution is the problem dealing with
the probability of evolution of life from the inception of the most primitive
life forms to intelligent beings. The various opinions here again cover a
wide spectrum, ranging from the extreme suggestion that the development of
intelligence is a single-valued consequence of the inception of life to less
categorical statements which regard the biological evolution as a succession
of critical, non-repeatable and unpredictable steps, a chain that can be
severed by the slightest of chances. If we adopt the latter point of view, the
life on Earth is a unique phenomenon, possible within the limits of the entire
Metagalaxy. This is clearly not a very appealing assumption.

The present author, because of total lack of background in biology, will
have to confine himself to an expression of hope that the fundamental problem
of exobiology will find its solution in the not too distant future and that the
probability of evolution of intelligence from primitive life forms is not too
low.

Finally we come to the remarkable problem of the evolution of intelligent
Societies outside the Earth, the problem of extraterrestrial civilizations.

As the emphasis here is on the evolution of society and we can essentially
regard the topic as falling within the framework of a new scientific discipline,
concerned with the study of hitherto undiscovered societies, we would use

the term 'exosociology" for this discipline, by analogy with exobiology.

I. S. Shklovskii's suggestion, "cosmosophy' /1/, is somewhat inconvenient in
our opinion and does not fully reflect the true task before us.

No science can be nourished by purely theoretical, "cosmosophic" con-
cepts, and exosociology is no exception to this rule. Experiments and
observations are essential components of any science. At the present stage
of its development, exosociology can draw for experimental data upon the
only civilization known to us, the Earth civilization. 'The real and significant
observational fabric of exosociology will be provided by the analysis of
signals from extraterrestrial civilizations, assuming that such signals will
be detected. It is this basic assumption that is reflected in the title of the
Introduction.

Exosociology is the subject of the present book. Exobiological topics,
i.e., problems relating to the origin and the evolution of life in outer space,

INTRODUCTION

are not considered. A fairly extensive literature is available at present
(see, e.g., /1/ and /2/).

The reader may naturally question the need and the urgency of a special
volume on exosociology at the present stage, when no systematic search for
signals from extraterrestrial civilizations has begun and the chances of
discovery of these signals are not very high. It is our belief, however, that
a book of this kind is urgently needed, and this for the following reasons.
First, systematic search for signals from extraterrestrial civilizations will
eventually be organized, and it is better to be prepared with all the necessary
theoretical and practical background information relating to this search.
Second, exosociological research may yield certain "byproducts" which will
be of considerable significance for "terrestrial" science. For example, the
search for radio sources of suspected artificial origin is entirely analogous
to certain problems of modern radio astronomy, and at first glance has no
relation to exosociology. The decoding of messages from outer space may
provide much valuable information relating to pure linguistic problems. And
so far we did not mention the forecasting of the future growth and develop-
ment of civilizations. Therefore, having organized a systematic search for
signals from extraterrestrial civilizations and proceeding with a research
into the various problems of exosociology, we will not end up losers even if
no extraterrestrial signals are detected in the near future. The potential
gain, on the other hand, is hardly imaginable.

The six chapters of the book deal with various aspects of the search for
signals from extraterrestrial civilizations. To help the reader, we will try
to present a general survey of the problem and the current view of its basic
aspects.

The theory of development of civilizations

The Earth civilization — the only known example of a society of intelligent
beings — has existed for a very brief period of time on the astronomical time
scale, for no more than a few millennia. The time interval accessible to
actual research is even smaller. And yet, the main topic of exosociology is
the study of civilizations over the entire span of their evolution, which, at
least in principle, may be comparable with the astronomical time scale
(millions and billions of years). In any case, signals can be detected only
from civilizations markedly exceeding the level of development of the Earth
civilization.

Exocosiology should thus be able to study supercivilizations, i.e., the
evolution of intelligent societies over very long, astronomical periods of
time.

It would seem that the solution of this problem should start with a detailed
forecast of the further growth of the Earth civilization. However, this
immediately leads us to a fundamental difficulty. Any forecast is essentially
based on an extrapolation of previous development. This extrapolation is
evidently valid over a period which is at most comparable to, and usually
much smaller than, the period of time on which the forecast is based. It is
not by chance that most forecasts of the future of mankind are limited to the
year 2000 (occasionally venturing to the year 2100)!

EXTRA TERRESTRIAL CIVILIZA TIONS

The intrinsic imperfection of the extrapolatory approach emerges from
the fact that its automatic application to the forecasting of the future develop-
ment of the Earth civilization inevitably leads to so-called "explosions" —
very rapid growth of some indices.

Probably the best known example is the "demographic explosion" or the
"population explosion," i.e., the conclusion that the Earth population will
become infinitely large around the years 2020—2030. Another example is
the "energy" or "power' explosion. Calculations show that around the
year 2100, the power production on the Earth will reach such a level that the
temperature of the planet will increase indefinitely. Finally, we seem to be
on the threshold of the so-called "information explosion," when the volume of
information accumulated by science will become infinite (this event is
"scheduled" for around the year 1980).

There is no doubt that none of these explosions will actually occur, but it
is not clear how the "critical" moments will be avoided and how the growth
characteristics will change to prevent the crisis. Repopulation of mankind
in outer space is often proposed as a universal remedy. A simpler
solution will probably present itself when the time is ripe. Analysis of the
succession of the growth characteristics is thus one of the principal problems
to be tackled in forecasting the future development of civilization (see
Chapter V).

Thus, despite the considerable interest attached to the forecasts of growth
of the Earth civilization, their contribution to exosociology is negligible. For
this reason, we will not go into these forecasts in any detail, and we would
only like to mention that according to A, Clarke /6/ and the forecasts
developed by the Rand Corporation inthe USA, the encounter with extraterres-
trial civilizations is deferred to the second half of the 21st century.

It therefore seems that at this stage it is more advisable to start looking
for general laws governing the development of intelligent societies and civili-
zations in some more abstract form, based on the modern cybernetic
concepts of complex systems. We should try to evolve general definitions of
the concept of civilization and to analyze the evolutionary trends emerging
from this system-theoretical definition. The following definition of a
civilization is advanced in Chapter I: "Ahighly stable state of matter capable
of acquisition, abstract analysis, and application of information for the
purpose of extracting the maximum quantity of information about the environ-
ment and itself and developing survival reactions." Chapter V mentions
another general feature: "Simple systems evaluate these outside stimuli only
in order to determine the state of the internal and the external media at the
materialtime, whereas more complex systems can respond to a forecast
future state of the environment as predicted on the basis of the current
measurements." Proceeding from these definitions, we can expect an
unlimited development of civilizations and an intrinsic tendency to establish
contact with one another. The cybernetic approach to the problem of super-
civilizations is discussed in more detail in Chapter VI.

We are not only very far from the solution of the fundamental problem of
exosociology, i.e., the elucidation of the general laws governing the develop-
ment of civilizations as intelligent societies, but we still have not formulated
this problem in precise terms. It is our belief, however, that the considera-
tions presented in Chapters I, VI, and partly V will help in this direction.

INTRODUCTION

Note that the establishment of contact with extraterrestrial civilizations
may not only lead to radical changes in our basic concepts regarding the
intelligent society, however "logical" these concepts had appeared prior to
the encounter with the other civilization, but also greatly affect the future
development of our own civilization. This will be the result of the "feedback
effect," often discussed, in particular, in connection with the beneficial or
harmful results of "interplanetary" encounters.

The search for signals from extraterrestrial
civilizations

Despite the tremendous volume of information accumulated by modern
astrophysics and radio astronomy, no such signals have been detected so
far. If we remember that most discoveries are quite accidental and happen
generally whenever they are least expected, there is no reason for over-
optimism in this respect. It is hard to say what the exact reasons are. It
may be that no other civilizations exist sufficiently close to the Sun which
are capable of sending signals into outer space. And yet, most authorities
are of the opinion that supercivilizations are quite abundant. We will be able
to reach sound conclusions, however, only after going through a
complete program of search for signals from other civilizations. This is
one of the reasons for our conviction that such a search program must be
launched immediately.

Incidentally, even if extraterrestrial civilizations do not send special
signals into space, there is a possibility that we will be able to "intercept"
their internal transmissions (television broadcasts, for instance). The
artificial radio emission of the Earth has reached by now a fairly high level
of intensity /1, 5/, and that of supercivilizations will be many times higher.
Combination of high-sensitivity receivers with large-base interferometers
(see below) will probably facilitate the problem of "interception" of the
transmissions of extraterrestrial civilizations.

The program of search for signals from extraterrestrial civilizations is
discussed in detail in Chapters I and III. The first step is apparently a
radio survey of the sky with the aim of detecting radio sources of minimum
angular dimensions. Indeed, the antennas of the sending supercivilizations,
irrespective of the particular information that they transmit, willbe very
small compared to the astronomical scale of distances. In principle, trans-
mitting systems of planetary size are possible, but even the planetary scale
is vanishingly small compared to the size of other radio sources in space.
The current resolving power of radio observations has reached 0'.005. This
resolution was attained with a radio interferometer using separate recording
in each arm. In principle, radio-interferometric observations can now be
made with a base of the order of the Earth's diameter, and in future the base
will probably be increased to about la.u. (giving resolution of 2. 107? angular
seconds! ).

There is a whole range of other criteria which identify the probable
artificial origin of a radio source. "These criteria are described and
discussed in detail in Chapters I and III, and a more general aspect of the
identification of artificial signals is given in Chapter VI. Regular variations
in the signal, definite polarization, and other features of this kind must be

EXTRATERRESTRIAL CIVILIZA TIONS

analyzed in great detail. The artificial nature of the signal can also be
inferred from the statistical properties of the electrical field of the radio
wave. The most reliable criterion, however, is nevertheless the exceeding-
ly small angular size of the source.

The choice of wavelengths at which artificial sources are to be sought
presents another important problem. It is generally agreed that the idea of
communication with extraterrestrial civilizations passed from the domain of
Science fiction to the domain of science in 1959, when Cocconi and Morrison
suggested that the signals of extraterrestrial civilizations should be sought
at the natural wavelength standard, the 21cm radio line of the hyperfine
Structure of atomic hydrogen. This suggestion naturally met with certain
opposition; in particular, it has been pointed out that the interstellar
medium is highly absorbing at this wavelength, so that the higher harmonics
of the 21cm line should probably be used.

There are, however, other natural wavelength standards, e.g., the radio
lines of the so-called A-doubling of the hydroxyl molecules OH. In fact,
four lines are observed, associated with the combination of A-doubling and
the hyperfine structure. The mean wavelength of the four lines is A= 18 cm.
For allthe four hydroxyllines, the interstellar absorption is significantly
lower than for hydrogen lines, but it is nevertheless quite high.

The hydroxyl radio lines have recently attracted considerable attention on
the part of radio astronomers and astrophysicists, following the discovery
of a "natural maser effect' at these wavelengths: very narrow (with a
Doppler width corresponding to a temperature profile of a few degrees
Kelvin) and very strong (with a brightness temperature of over 10!? deg)
highly polarized hydroxyl lines have been observed for a number of sources
located near the regions of hot ionized interstellar hydrogen. 'The unusual
behavior of these lines explains their new name, the "mysterium lines." If
we further remember that the radio sources of "mysterium" lines are
characterized by the smallest known angular dimensions, of the order of a
few thousandths of an angular second (this corresponds to linear dimensions
of a few astronomical units for their distances from the Earth), no wonder
that these sources are suspected as being of artificial origin.

We are far from suggesting that the mysterium" sources are extra-
terrestrial civilizations, but this example clearly illustrates the great
importance of detailed observations and analysis of all the "suspicious"
objects.

Further note that at centimeter and decimeter wavelengths, which are
the most suitable for purposes of interstellar radio communication (the
interstellar noise is the least at these wavelengths, see Chapters 1 and II),
there are other molecular lines which in principle can be used for signal
transmission by extraterrestrial civilizations. Finally, radio transmission
is also possible and even highly probable in the continuous spectrum between
10 and 50cm wavelengths, and this wide frequency band ensures a sufficient-
ly high rate of information transmission (Chapters I and III).

Recently considerable attention has been attracted by the discovery, on
6 August 1967, of the so-called "pulsars," pulsating radio sources with a
remarkably regular periodicity of pulse repetition in a continuous spectrum.

The observations of pulsars in the first months following their discovery
was closely linked with the problem of search for signals from extraterres-
trial civilizations. We will therefore consider this chapter of science in

INTRODUCTION

some detail, The name pulsars was assigned to certain objects which emit
discrete and very short pulses (with a duration of the order of a few
hundredths and even thousandths of a second) in a wide region of the
continuous radio spectrum. In the intervals between the successive pulses,
no pulsar emission has been observed so far. The radio pulses differ in
shape and in amplitude, i.e., in emitted radio power. The pulses reveal

a certain fine structure: those of numerous pulsars are made up of so-called
subpulses. The pulses of different pulsars have different shapes, and even
the pulses of one pulsar are variable in this respect. The magnitude of
pulsars is variable between even wider limits, and occasionally they vanish
altogether. Inmany, though not inall, cases, the pulsar pulses are polarized.
At least some of the pulsars probably emit pulsed radiation in the visible
Spectrum also. The various features described so far are quite usual for
natural astrophysical sources, and possibly even for ordinary stars. Certain
features of the pulsar radio emission are quite similar to the sporadic radio
emission of the Sun. However, one of the pulsar properties — in fact, their
main property which is responsible for their very name — appeared highly
unusual. The pulses revealed a strikingly regular periodicity of recurrence.
The first of the discovered pulsars showed pulse recurrence periods close to
l sec, and the exact period of each pulsar remained constant with astonishing
precision: over a year, the period did not change to the seventh or eighth
position after the decimal point. For example, the period of the best known
pulsar, CP 19019, is 1.33730109 £1075sec. Soon after that, it was estab-
lished that the pulsar periods systematically increase (the change is in the
Seventh significant digit during one year). This strict periodicity led
A.Hewish, who headed the group responsible for the discovery of pulsars

in Cambridge (England), to the suggestion of the possible artificial origin

of pulsars. The press at that time succinctly described the pulsars as the
signals of the "little green men." A. Hewish kept the discovery as a closely
guarded secret for about six months after the observation of the first pulsar,
a highly unusual development inthe modern scientific community. It was only
after the discovery of three other pulsars in Cambridge that the results
were announced. Almost simultaneous discovery of several extraterrestrial
civilizations is a highly unlikely event.

Note that the existence of a strict periodicity in natural processes which
take place in astronomical objects is by no means an unusual phenomenon,
Obvious examples are the axial rotation periods of planets or binaries.
Certain variable stars (the relatively small group of RR Lyrae stars, typical
type I population stars) are distinguished by exceptional stability of light
variation: their periods do not change significantly over a million cycles.

So far, however, the astronomers have dealt with periods measured in
hours and days, whereas in pulsars the characteristic periods are seconds
or fractions of a second, but this does not appear to be a fundamental
distinction.

Besides strict periodicity, the pulsars show nothing that supports the
hypothesis of artificial origin (see Chapters I and III). This hypothesis
survived for a few months only. By the end of 1968, 27 pulsars had been
discovered with periods ranging from 300 to 3 seconds. The properties of
pulsars proved to be highly interesting and highly unusual: some theories
identify these objects with spinning neutron stars (these theories explain both
the strict periodicity and the increase in period), However, the pulsars can
be said to definitely fall outside the scope of our book.

EXTRATERRESTRIAL CIVILIZA TIONS

The modern theory of communication enables us to analyze the conditions
of signal transmission through interstellar space, to consider the requirements
to be met by the transmitting and, especially, the receiving systems and
antennas. This analysis, carried out in considerable detail in Chapter III,
will help to select the optimum antenna parameters, receiver band widths,
and scanning periods in connection with the program of search for extra-
terrestrial signals. We would only like to stress that the main problem falls
into two separate parts: the direct search for signals (''discovery of artificial
sources") and reception of information from extraterrestrial civilizations,
For straightforward detection purposes, the useful signal may be much
weaker than the noise level, These signals can be picked up with the aid of
averaging techniques (as is often done in radio astronomy), but part of the
information is naturally lost in the process. If we are interested in merely
detecting signals from extraterrestrial civilizations, without interpreting
their meaning, the "power" of the civilizations may be several orders of
magnitude less than in cases when full reception of information is required
(and the maximum distances are correspondingly larger). This means,
incidentally, that the first instances of signal detection from extraterrestrial
civilizations will not lead to catastrophic consequences.

We do not intend to present here any specific programs of search for
extraterrestrial civilizations. The actual program will be decided upon only
after a comprehensive and all-sided analysis of the possibilities of modern
radio-astronomical equipment, taking into consideration the actual observa-
tion time available on the largest radio telescopes for this project. The use
of radio interferometers with a base comparable to the Earth's diameter will
be impossible without close international cooperation on the project.

The authors nevertheless hope that the analysis of the problem of search
for signals from extraterrestrial civilizations, presented in this book, will
promote the development of a large and comprehensive program with higher
chances of success than the well-known Ozma project initiated by F. Drake
in 1956 for detailed observations of the two close neighbors of the Sun,
£ Eridani and t Ceti.

Decoding aspects of the program of search
for extraterrestrial civilizations

Before any signals have been received, we are in no position to discuss
their probable information content. There is absolutely no point intrying to
guess now whether these will be television images (the most comprehensible
language, at least from our point of view) or messages based on the princi-
ples of formal logic, akin to the famous LINCOS language, or perhaps
something entirely different.

It nevertheless seems that we are ripe for a precise formulation of
certain basic problems relating to the decoding of unknown messages.
Consider one example. Suppose a certain message has been received; let
this be a text written in an unknown language, with an unknown alphabet and
unknown rules for division into sentences and words; even the letter codes
are unknown. The only available piece of information is that we have
received a sequence of signals, e.g., pulses, of definitely artificial origin.
Can this text be decoded so as to disclose its meaning and contents? For

INTRODUCTION

purposes of decoding, it is necessary (though not sufficient) to determine the
letter codes and the division into words and sentences, to establish the
grammar of the language, to compile a dictionary, and to elucidate the
pronunciation of the letters and the words.

Consider another example. A fragmentary message (e.g., distorted by
noise) has been received, but it is almost certainly a part of an image (a
static television picture). Can we reconstruct the entire picture from the
received message, i.e., determine the number of lines and scanning
elements in each line? The best-known example of messages of this kind is
Drake's cosmogram (described in Chapter IV), in which a sequence of
1271 elements (ones and zeros) is used to code the picture of certain crea-
tures (remarkably like human beings, only somewhat taller) inhabiting the
fourth planet of some planetary system. The deciphering of this cosmogram
is greatly facilitated by the fact that the number 1271 can be split either into
31 lines of 41 elements each, or into 41 lines of 31 elements each. There
are thus two alternative solutions, and the right answer is almost obvious.
However, if we miss a few of the first elements of the message, the screen
is no longer rectangular and the message will probably be undecipherable.

There is, of course, a possibility that the signals from extraterrestrial
civilizations contain the key for the decoding of the transmitted message.
The question is directly related to the topic of call signals, which should
identify the artificial origin of the signals. This idea opens wide horizons
for various assumptions and speculations. We will consider the problem of
call signals and simple keys for decoding in Chapters I, III, IV, and VI. In
our opinion, however, it is better and more worthwhile to concentrate on the
problem of decoding of unknown messages assuming total absence of any
decoding keys. This constitutes the topic of Chapter IV, which was written
by a professional linguist.

The method of decoding proposed in this book essentially amounts to what
is known in physics as the method of construction of correlation functions
(they are called quality functions in Chapter IV) for messages. Indeed,
certain combination rules exist for the consonants and the vowels, for words
which belong to different grammatical classes, and correlation functions
constructed for different symbols of the received message therefore provides
certain identifying information about these symbols. If the message com-
prises the scanning elements of a picture, the correlation function permits
reconstructing the successive lines andthenthe entire picture. This decoding
procedure naturally involves a large volume of computations, and therefore
it must be handled by computers. The problem of decoding thus reduces to a
construction of an algorithm for the computation of correlation functions and
their comparison with certain criteria (of the type of the entropy criterion)
which make it possible to select the best solution (the entropy of ordered
distributions is minimum). It is moreover clear that since the decoding
procedure is based on statistical processing, a sufficiently large sample,
i.e., a sufficiently long message, is needed for the decoding to prove
effective in complex cases. Simple examples nevertheless can be solved
using short messages.

We would like to stress that Chapter IV mainly deals with the decoding of
messages from the linguist's point of view. The reader interested in the
general principles of decoding may read only the first seven sections. The
remaining four sections contain various algorithms intended for the solution

EXTRA TERRESTRIAL CIVILIZATIONS

of more complex problems. Despite the sophisticated algorithms, however,
we are still very far from complete decoding of long texts in an unknown
language. Yet the principles have crystallized, and the rest is a technical
matter.

* * *

We tried to present a brief survey of a new scientific discipline — exo-
Sociology, the search for signals from extraterrestrial civilizations — and
at the same time review the contents of this book. I would now like to add
a few comments in my capacity as the editor of this volume.

The original intention was that each chapter should embrace one well-
defined aspect of the problem of search for signals from extraterrestrial
civilizations. The result is thus not a collection of papers, but a kind of
monograph. The main difficulty, however, is that exosociology, like any
new Scientific discipline, still gropes uncertainly among differences of
opinion and lack of firmly established concepts. Even the different contri-
butors to this volume differ in their opinion on certain subjects. It was not
the editor's intention to impose his own point of view upon the authors or to
act as an arbitrator. As a result, however, a number of topics, e.g., the
concept of a civilization, thedate ofthe energy explosion, etc., are discussed
in different chapters, sometimes from different points of view. The reader
will have to decide for himself whose arguments sound the most convincing.
He may even feel free to form his own opinion on the subject.

It should be emphasized, however, that these "differences of opinion" are
relatively few and, on the whole, the contributors have pursued the original
aim, namely a scientific discussion of the problem of search for signals
from extraterrestrial civilizations on the modern level, in order to stimulate
further interest in this problem.

The book is intended for a wide audience, although it is not a popular book
in the usual sense of this word The authors did their best to maintain a high
Scientific level in their presentation, without going into tedious technical
details which are of interest to narrow specialists only (the only exception
to this rule is probably the second part of Chapter IV). The main difficulty
for the reader is the great variety of subjects covered: radio astronomy,
theory of information, linguistics, cybernetics, aspects of civilization...

Some readers will probably feel that certain sections are much too
superficial, whereas others are excessively detailed. Certain chapters are
too simplified, and others are too complicated. In partial justification of
this, we would like to point out that it is very difficult to maintain a consis-
tently uniform level of presentation in a volume written by a team of
contributors on such a wide spectrum of subjects.

The present book is radically different from previous publications on the
subject of extraterrestrial civilizations. References /3/ and /4/, for
example, are collections of articles and papers, and therefore do not provide
a comprehensive picture of the problem. Moreover, they are largely out-
dated by now.

W.Sullivan's book is more of a popular discussion of the various events
associated with the problems of exobiology, and thus does not provide a
consistent analysis of the fundamental problems.

INTRODUCTION

I. S. Shklovskii's book /1/ is unquestionably of the greatest interest.
Unfortunately, it was written quite a number of years ago and numerous
aspects of the problem of extraterrestrial civilizations are therefore not
mentioned, Furthermore, the presentation is much more popularized than
in the present volume.

It would seem that the present volume is the first scientific monograph
in the literature on the subject of search for signals from extraterrestrial
civilizations.

In conclusion, all the contributors would like to acknowledge the great
help of Acad. V. A. Kotel'nikov for valuable suggestions that helped to
improve the finished product, and especially the assistance of L. I. Gudzenko,
who read through the entire manuscript and offered numerous comments
concerning the general presentation and the particular problems discussed.

Bibliography

1. Shklovskii,I.S. Vselennaya, zhizn', razum (Life and Intelligence in
the Universe). 2nd Ed.— "Nauka." 1965.

2. Sullivan, W. We are not Alone. — McGraw-Hill. 1966.

3. Cameron, A. (Editor). Interstellar Communication. — New York.
Benjamin. 1963.

4. Vnezemnye tsivilizatsii (Extraterrestrial Civilizations). Proceedings
of a Conference.* Byurakan, 20—23 May 1964.— Izd. AN Arm.
SSR. 1965.

5. Kaplan,S.A. Elementarnaya radioastronomiya (Elements of Radio
Astronomy). — "Nauka." 1966.

6. Clarke,A.C. Profiles of the Future. — Harper and Row. 1962.

* [English translation published by Israel Program for Scientific Translations, Jerusalem, IPST Cat.No.1823,
NASA TT F-438 TT 67-51373.]

Chaptev I

THE ASTROPHYSICAL ASPECT OF
THE SEARCH FOR SIGNALS FROM
EXTRATERRESTRIAL CIVILIZATIONS

$1. INTRODUCTION

The search for extraterrestrial civilizations is intimately linked with the
principal problems of modern astrophysics. Let us try to establish what
part of the proposed search program actually coincides with astrophysical
research and what the specific requirements of the observations in this
program are.

Accurate long-range prediction of the principal problems and the direc-
tions of development of space science is a fairly difficult problem. The
current tendencies, however, which will leave their indelible imprint on
the next few years are quite obvious.

In the next 5 — 10 years, all the radiation sources
with the largest observable flux in every region of the
electromagnetic spectrum will have been discovered
and studied to a certain extent (A).* This is a realistic goal
in view of the development of electromagnetic radiation detectors, i.e.,
radio receivers, bolometers, photosensitive detectors and materials,
and photon counters. The sensitivity of these devices will soon reach
the natural limit (in some spectral regions, this limit sensitivity has
been attained already, e.g., the modern photon counters used in mea-
surements of X-ray radiation from outer space detect every single
impinging quantum). When the limit sensitivity is attained, we will be
able to cover various cosmic objects in the entire electromagnetic
Spectrum, and thus virtually all thé astrophysical information con-
tained in cosmic radiation. We are thus nearing the solution of a highly
important astrophysical problem:

Identification and exploration of the main (in terms
of some parameter) cosmic objects (primarily objects
of maximum luminosity, or radiation power, in a given
Spectral range, objects of the largest mass, and objects
which account for the bulk of matter in the Universe )(B).

The primary problem of this exploratory trend is the determination of
the luminosity function N,(L,) and the mass function Ny(M) of all the objects,
where /, is the spectral power radiated by the object. Unfortunately, the
solution of problem A does not imply a simultaneous solution of problem B

* The main propositions of this chapter are identified by bold -face letters.

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

(although the inverse is probably true).* Indeed, objects of the highest
luminosity (e.g., supernovae, quasars) are exceptionally rare in the
Universe. Therefore, the mean distance between these objects (and
hence the most probable distance to the nearest source) is tremendous,
and these high-luminosity objects may not be the brightest. The nearest
quasar 3C 273 has a brightness of 12.5 stellar magnitudes in the optical
spectrum. There are over four million stars of this magnitude in the
sky, and the quasar therefore escaped optical detection for a long time;
in the radio spectrum, on the other hand, this quasar is one of the
hundred brightest objects, and the radio astronomers noticed it
immediately.

Let us estimate the observability in a given spectral range using the
luminosity function. Let

N, e Lv,
where the index n can be determined from observations. The number of

sources in unit volume with luminosities between ib and iL, is then
given by

Nyx De
The observed flux from the nearest source whose luminosity falls between

41, and $L, is

L
Fy Rr
where the mean distance between the source is Rœ N^.
Hence,
5-2n
Fi «xl, , (1.1)

We see from this relation that if n>5/2, then Sa? <0, and the lowest

luminosity sources prevail among the sources with the maximum observed
flux; if, on the other hand, n< 5/2, the situation is reversed, and the
maximum luminosity sources prevail among the brightest objects.

The first of the two possibilities obtains in the comparison of the mean
radiation from normal galaxies, radio galaxies, and quasars. These
objects are not numerous, so that nis high, and therefore the normal
galaxies prove to be the brightest among all the extragalactic optical
sources.

On the other hand, if w consider the optical radiation of normal
galaxies only, we have n<5/2 and therefore the brightest observed
objects are the most powerful. A similar situation is observed for
extragalactic radio sources. Figure 1 plots the radio luminosity func-
tion /46/, which shows that in a wide range of luminosities, n~ 2.2,
so that the brightest objects are also the most powerful, and it is these
powerful sources that are mainly studied today. Unfortunately, no such
analysis can be undertaken for the mass function, since no reliable data
are available at this stagc. The only established fact is the mass

* Wy and Ny is the number of objccts in unit volume with radiation power in the range L +6 and mass

in the range M+6M, respectively, where 20Ly and 26M are unit intervals of luminosity and mass.

EXTRATERRESTRIAL CIVILIZATIONS

distribution of the stars /47/. This distribution is also adequately fitted
with a power function with n £2.35.

A, Mpc^*

Normal
galaxies

Radio
galaxies

m*
m
Quasars
m
m" 0% w” wË m”
donis.
V Hz . ster

FIGURE 1. The luminosity function of extra-
galactic radio sources.

Problem B stresses the main tendency of development of the astro-
physical research in the near future. In particular, if the activity of
extraterrestrial civilizations is responsible for the radiation power of
some astronomical objects, these civilizations stand a good chance of
being discovered. Since in the nearest future all the regions of
the electromagnetic spectrum will be accessible to space exploration,
we have to prepare a suitable research program and to assess the chances
of success in our search for extraterrestrial civilizations.

$2. THE MAIN DILEMMA

The main starting point for our problem probably stems from the following
dilemma:

There is a high probability that civilization is a
universal phenomenon, and yet there are no currently
observed signs of cosmic activity of intelligent
creatures (C).

Indeed, the data available on the number of planetary systems and the
conditions for the evolution of life on planets suggest that life is probably
a fairly commonplace and regular occurrence in the Universe. A detailed
analysis of these topics will be found in /1,2,3/.*
^ In particular, recent paleontological data convincingly prove that the inception of life on the Earth some

3 billion years ago took place simultaneously in numerous independent channels /4/.

I. ASTROPHYSICAL ASPECT OFSEARCH FOR SIGNALS

According to most estimates, the age of our planetary system and the
age of the Sun (reckoned from the time of their condensation) is from 4
to 6 billion years. It is significant that both the Sun and the planetary
System are second-generation objects, but since the age of the oldest
objects in the observable part of the expanding Universe (or, more pre-
cisely, the age of the first-generation objects) is at least 10 billion
years, there are probably planetary systems billions of years older
than the solar system. This conclusion suggests the possible existence
of civilizations which are billions of years more advanced than our
civilization. Taking into account the present rate of progress of our
civilization, we can probably expect something nearing intentional and
controlled reorganization of all matter in our part of the Universe from
civilizations developing over these cosmogonic periods.

And yet, our astronomical data at first glance do not provide any
indications of such cosmic activity. In our opinion, a detailed analysis
of proposition C may provide the best foundation for the discussion of
the program of search for extraterrestrial civilizations (EC). We will
try to evaluate the various aspects of this dilemma in order to critically
assess its relevance.

There may be two alternative answers resolving the dilemma:

1) either the current data on the absence of "supercivilizations'" are
wrong;

2) or there exists some fundamental factor slowing down the develop-
ment of each and every civilization.

$3. THE COMPLETENESS AND RELIABILITY OF
MODERN ASTROPHYSICAL DATA

As we have already noted, there can be no serious doubt regarding the
existence of numerous planetary systems (although planets with masses
of the order of the Earth's mass cannot be directly observed with
modern telescopes (see /1,2,3/). Estimates of the number of planets
which may be suitable for the evolution of life do not give any indication
of the Earth's unique position in the Universe, either (see /1,2,3/).

The Sun and the solar system are thought to be second-generation
objects, but if it were not so, there would be a definite probability of
the Earth being the oldest object of this kind in the observable part of the
Universe and our civilization being also the oldest.

At this point, we will have to review the current evidence relating to
the age of the solar system.

Most stars whose physical parameters are close to those of the Sun
remain in a steady-state condition for a long time, retaining constant
radius and luminosity. The loss of radiant energy is made up by the
energy released in nuclear reactions in the stellar interior. These con-
cepts were used to develop the theory of stellar evolution according to
which the steady-state phase of the Sun's evolution may take about
13 billion years, i.e., the entire evolutionary phase of the Metagalaxy.
On the other hand, the age of terrestrial rocks and meteorites deter-
mined by chemical analysis of radioactive isotopes and decay products

EXTRA TERRESTRIAL CIVILIZA TIONS

is 4—5 billion years. This figure is usually adopted as the age of our
planetary system and the Sun, since the modern theory of formation of
planetary systems points to simultaneous condensation of the planets
and the primary star from interstellar gas-dust clouds.

Recent results, however, seem to have substantially revised upward
the age of the Farth and meteorites (see /5/). Thus, Fisher /5/ reported
the results of K— Ar dating which gave an age of up to 10 billion years for
some iron meteorites. The same technique gave an age of up to 10.8 billion
years for terrestrial rocks /6/. Although these and other similar data by
no means provide a conclusive proof of a new longer evolutionary scale of
our planetary system, we cannot just ignore them.

Another aspect of this problem is related to the chemical composition
of the planets. The condensation of Earth-type planets requires a sufficient
content of the heavy elements in the interstellar medium, and we are thus
faced with the unanswered question of the evolution of the interstellar
medium and the genesis of the heavy elements in general.

In accordance with modern data on the evolution of the observable part
of the Universe, it seems that all the chemical elements were formed in
nuclear reactions from an original pure hydrogen plasma. Until] recently,
these processes were assumed to take place in stellar interiors only, the
heavy elements being produced by reactions during supernova explosions.
Subsequently, the heavy elements are ejected into the interstellar medium
/1/. This mechanism obviously supports the hypothesis which treats the
Earth-like planets as second-generation objects.

Lately, however, a new class of first-generation objects were dis-
covered, which also show a high content of heavy elements. We mean
here the quasars. The objects are primarily remarkable in that
their radiation power is the highest among all the known sources of radia-
tion in the Universe. As a result, they can be observed over tremendous
distances and, because of the finite velocity of light, they provide a tool
for probing into the distant past of the Universe. Figure 2 is a photograph
of one of the farthest quasars 3C 9. The spectral lines of these objects
show a strong red shift because of the observed expansion of the Universe.

For 3C 9 the red shift is z = the 2, so that all the wavelengths increase
relative to the laboratory standards by a factor of 1 + zap = 3. The time

between the emission and the observation of radiation for distant objects
essentially depends on the particular cosmological object used. In the
Einstein— de Sitter model (space curvature k = 0, acceleration parameter
qo = Y2), the propagation time of a light signal is

2.2 (l+z)"e-1
"T8. +a% (1.2)

Here H, is the Hubble constant (for small red shifts z, the distance to the

object is qm The value of this constant is Hy — 30 km/sec - 10? light years.
In this model, the light from 3C 9 takes about t= 5.3 billion years to reach
the Earth. (The relevant data for the calculations using other models will

be found in /7/.)

I, ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

o® -.
N > 8
` t *
i *
"m : M à
dd * : .* e
*
Ld e
s * s
"e
V . . @ +
a
¥ a
*
s * * *
B - e ICI
Sa eed . »
E
P os .
iUt.
* . a
t * s i e
$ $ BD
€

LE IS Lr vu ds
FIGURE 2. The quasar 3C 9.

The crucial point of the entire problem, as we have noted before, is
the discovery of normal chemical composition in these objects /8/. In
other words, the mean abundance of the chemical elements (at least of
the most abundant species) in quasars is close to that observed in the
neighborhood of the Sun. At the same time it has been established that
quasars lie in regions where the concentration of ordinary galaxies is
much below the average (between clusters of galaxies). They apparently
form directly from the intergalactic medium. The heavy chemical ele-
ments are possibly synthesized in the quasar interiors, since the con-
ditons prevailing in quasar explosions are probably even more favorable
for nucleogenesis than supernova explosions. However, the similarity
in the chemical composition of various quasars is really striking. It
is therefore not improbable that the heavy elements were synthesized at
an even earlier stage of evolution of the Universe, and the intergalactic
medium from which the quasars form have the same composition as the
interstellar medium. Thus, the age of the heavy elements
needed for the formation of Earth-type planets may
be comparable with the age of the observable part of
the Universe.

EXTRA TERRESTRIAL CIVILIZA TIONS

The above new data point to the possible existence of planetary systems
whose age is close to the age of the oldest objects in the Universe. However,
the best evidence that the Earth is not the oldest planet is provided by
certain observations as interpreted in the light of the modern theory of
stellar evolution. As we have noted before, stars after condensing from
the interstellar medium remain in a quasistationary equilibrium for a
long time, and the radiant energy losses are balanced by the nuclear
reactions in the stellar interior. The length of this phase increases and
the luminosity decreases with the decrease of the stellar mass. When
the hydrogen has been "burnt up," the stellar nucleus compresses, its
temperature increases, and the stellar radius increases. The stars of
various masses in which an equilibrium is maintained by thermonuclear
fusion reactions (mainly producing helium) constitute the so-called
main sequence. Stars which have exhausted their hydrogen supply
move from the main sequence to the group of red giants. The duration
of the main-sequence phase in the life of a star and the presence of red
giants in some group of stars clearly
make it possible to find the age of
that group. Figure 3 shows the so-
called Hertzsprung — Russell diagram
for 11 star clusters. The horizontal
axis gives the color of the star (the
difference between the photographic
and the visual stellar magnitudes),
and the vertical axis marks the
absolute visual stellar magnitude.
The envelope on the left is the main
Sequence curve, and it also plots
the color and luminosity distribution
of the stars in the youngest of the
11 star clusters, NGC 2362. The
-44 4 G6 46 L2 16 80 vertical axis on the right gives the

ae age corresponding to the duration of
FIGURE 3, Hertzsprung-Russell diagram for the main-sequence phase of a star
some star clusters. of a given luminosity. The arrow
marks the position of the Sun on the
main sequence. The curves branching
off the main sequence in the upward right direction plot the color and
luminosity distribution of the red giants in each cluster. The branching
point evidently gives the age of the cluster. We see from Figure 3 that
the branching point of NGC 188 lies below the Sun, which indicates that
the age of this cluster is higher than the time that the Sun has so far
spent on the main sequence. This conclusion is also borne out by some
other data. According to its position in the Galaxy and its velocity rela-
tive to the galactic center, the Sun belongs to the disk-type or the inter-
mediate stellar population, which are all characteristic second-generation
objects. First-generation objects (the halo subsystem) which formed
originally when the galaxies condensed eject gases enriched with heavy
elements. These gases are mixed with the leftover interstellar gas of
the first condensation, settle to the plane of rotation of the galaxy, and
condense into the stars of the disk and the intermediate subsystems—

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

the second-generation stars. Spectroscopic observations of the Sun
reveal an abundance of heavy elements characteristic of second-generation
objects.

Let us now consider the second component of our dilemma, namely
that no signs of activity of supercivilizations have been discovered so
far. What is the supporting astrophysical evidence in this respect?

Let us try to estimate the percentage of the currently available astro-
physical information out of the total quantity of information which may
be contained in the entire electromagnetic spectrum.

Modern astrophysics yields a surprising wealth and variety of infor-
mation. Optical and radio catalogues list thousands of stars, galaxies,
and nebulae. For many of these objects, chemical composition and the
physical state of matter are known. The current hypotheses regarding
their evolution show a satisfactory fit with the results of statistical
analysis of observation data. The observational tools of astronomy have
become so effective that radiation sources can be explored at distances
of billions of light years. This profusion of data may create the impres-
sion that the current hypotheses give a consistent picture of the evolution
of the Universe, that almost all the main objects in the Universe have
been discovered, and that it only remains to clarify a few minor details.

In my opinion, this is a basically erroneous attitude, although the
state of the observational art is such that the structure of the Universe
will be elucidated in general outline in the nearest future.

There are numerous examples of outstanding discoveries in astro-
physics which were made in recent years only (e.g., the discovery of
quasars, the background relic radiation, which accounts for a substantial
fraction of the total electromagnetic radiation, molecular generation of
the 18-cm hydroxylline, pulsars). Some of these recently discovered
objects may prove to have an immediate bearing on our search for
supercivilizations. On the other hand, our knowledge of the quantity
and state of solid matter in the Universe is negligible.

As we have noted at the beginning of this chapter, sources of qualita-
tively new information about cosmic objects may soon become available
with the mastering of new frequency regions. What percentage of the
entire frequency spectrum have we mastered so far? The search for
the main radiation sources in each frequency range is far from pro-
viding a complete coverage of all the sources of information, but even
this basic problem has not been solved so far. The percentage of the
mastered frequencies can therefore be regarded as an upper bound
estimate of the available quantity of information. It is in this sense
that we should interpret the concept "mastered frequency range."

A frequency range is said to have been mastered if more than 30%
of the total sky area has been scanned for sources at a given wavelength,
and more than 100 cosmic objects have been discovered as a result of
this search. We have to distinguish between two cases:

1. The search for objects emitting a wide spectrum of frequencies
(spectrum width Av~v).

2. The search for objects emitting in narrow spectral lines.

The second problem is clearly incomparably more complex than the
first, since it involves coverage of the entire electromagnetic spectrum

EXTRATERRESTRIAL CIVILIZATIONS

with narrow-band filters. Roughly speaking, the number of measurements
required in case 2 isafactor of 3. greater than in case 1. For the radio

lines of the interstellar hydroxyl OH at 18-cm wavelength we have for

some objects Ay = 3-108,

This narrow-band scanning is of the greatest importance both for
astrophysics and for the search for civilizations. So far, however, no
narrow-band survey of the sky has been carried out either in the optical
or the radio spectrum (the only possible exception is the complete survey
of the sky in the interstellar hydrogen line }= 21 cm in a band of about
1 MHz with spectral resolution of about 10kHz). The percentage of the
available information on pure monochromatic sources is therefore still
exceedingly small.

The search for wide-band sources is a much simpler problem. The
number of mastered frequency ranges (e.g., octaves) for these sources

is determined by the expression / «In I, where v, and v» are the minimum
1

and the maximum frequency of the survey. The percentage of the mastered
frequencies is clearly given by

ü Cae T tm (1.3)

where (3>) , is the maximum to minimum survey frequency in the radio
i jra
spectrum, (2) ditto in the optical spectrum, and the ratio B in the
V1 7 opt
denominator is determined by the maximum and the minimum frequencies
of astronomical surveys in future.
At present, radio surveys have been conducted at frequencies between

40 and 400 MHz, so that (32)... — 10.
In the optical spectrum, photographs and observations of individual
c +
sources covered the range from 3000 to 6000 A, i.e., (2). =2.

In the other frequency ranges, there are only isolated observations of
small sky areas, which constitute a negligible percentage of the entire
quantity information.

What is the value of the denominator in (1.3)? The low-frequency limit
of astrophysical observations has been fixed with fair certainty. The
minimum frequency is w~ 1 MHz, since at lower frequencies the inter-
stellar medium is opaque and only objects very close to our planetary
system can be observed.

The high-frequency limit is more difficult to determine, and it apparently
linked with the quantum nature of electromagnetic radiation. As the fre-
quency increases, the energy of each detected quantum becomes higher.
Now, as the energy resources of astronomical objects are limited, the
number of quanta reaching the detector decreases as the quantum energy
increases. A more detailed estimate of the frequency v: will be given

I, ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

later on. For the time being we take vz~ 10!? Hz (wavelength 3A), Then
I = 10!?, and the percentage of mastered frequencies is
ei 1g10+ 1g2

i.e., even in the relatively easy search for wide-frequency bands, we have
so far mastered a low percentage of the total information available. Note
that of the 89% of the missing information, 42% falls between 10? and

10 Hz (centimeter, millimeter, submillimeter, and infrared waves)

and 25% between 10!9 and 101? Hz (ultraviolet radiation and X-rays).

The limits vı and v? of the entire electromagnetic spectrum are fixed
with considerable uncertainty. We have probably underestimated its
width, so that the 1196 is an overestimate.

Let us now estimate the number of cosmic sources which can be dis-
covered in a given electromagnetic frequency range. As we have noted
before, the sensitivity of some radiation detectors has now almost reached
the physical limit determined by the quantum nature of the electromagnetic
radiation and the background of cosmic radiation. Therefore, the success
of a search for sources of small angular dimensions will depend on the
number of quanta per unit detector surface area and the possibility of
resolving the various sources. Y

On the long- wave side the number of sources is limited by the angular
resolution of the antenna. The number of antenna beam widths accommodated
by the celestial sphere is

= 11%, (1.4)

A 2
N EAT IX i (1.5)

where A, is the effective collecting area of the telescope. In the radio
spectrum, the best antennas have Aœ À?, This is so because the relative
precision with which a reflecting surface can be manufactured is approxi-

mately constant, i.e., T ~ const, where D is the reflector diameter, and

e is the mean error surface; for a reflector to be effective in a given
frequency range, we should have e<0,14. Thus, in the radio range, the
maximum number of distinguishable sources N, is independent of wavelength.
A survey of the hundred brightest sources in every frequency range
clearly does not require antennas of maximum capacity. Nevertheless,
taking W^ 100, we should change the effective area A,«4? on passing
from one frequency range to another.
Relation (1.5) leads to an important conclusion. When working with
the instrument of maximum capacity and when surveying different frequency
regions for a constant number of the brightest sources, the expected

quantity of information is proportional to In ~ and the above estimates
based on (1.3) remain valid. i

For short-wave observations (K-ray and gamma-ray frequencies), we
can work with equipment counting every single incoming quantum and
faithfully indicating the direction from which it arrived. The number of
sources that can be discovered in a time « therefore cannot exceed the
number of quanta from these sources which reached the detector,

Ni g PAE, (1.6)

EXTRA TERRESTRIAL CIVILIZA TIONS

Here pis the total density of electromagnetic radiation in a given frequency
range in unit volume from all the sources. According to measurements at
wavelengths shorter than the optical spectrum p<10-! erg/cm?. The
parameter A, (e.g., the cross section of the gamma counters) hardly
changes with wavelength in this case, and therefore N, diminishes as
the frequency increases. Clearly, the frequency at which N,—N;is that
particular v, above which only a negligible fraction of information is
contained. (A, cannot be increased with increasing frequency because
of formidable technical difficulties.)

Thus, equating (1.5) and (1.6) and assuming A, to be of the same order,
we find

"(E (1.7)

Because of the weak dependence of v: on the particular values of the para-
meters, we may take p< 10°’ erg/cm?, survey time t~1 year ~3-10’ sec,
and this gives v; « 5-10" Hz.

Let us briefly reiterate the conclusions which follow from the above
discussion: despite the great advances in astrophysics, our information
is still insufficient to disprove the possible existence of supercivilizations
by arguing that so far no signs of their activity have been observed. Ata
later stage we will consider the possibility that some of the already known
objects (e.g., quasars) are in fact products of activity of supercivilizations.
On the other hand, the astrophysical data firmly indicate the existence of
planetary systems much older than the solar system. This provides justi-
fication for setting up a detailed program of search for extraterrestrial
civilizations.

We have considered some of the astrophysical aspects of the fundamental
dilemma (C) and our conclusion is that the entire dilemma is most probably
a product of insufficient knowledge on our part. If this is indeed so, we
must try to establish what astrophysical signs the activity of superciviliza-
tions can be expected to produce. This problem probably can be solved by
analyzing some general features of the development of civilizations over
cosmogonic periodics, It should be clearly understood that our knowledge
in this field is pitiful, On the other hand, we will not be able to go any
further without making some basic assumptions. There is no doubt that the
laws governing any field of activity of our civilization can and should be
formalized and systematized to a certain extent. This approach will
probably prove helpful in our analysis also. Some general considerations
on this subject are given in the next section.*

$4. CIVILIZATIONS AND THE MAIN FEATURES
OF THEIR DEVELOPMENT

We are primarily concerned with the highest level of development and
the general trend of activity of civilizations which we can expect in the
initial phases of the search program. Once these preliminary points are
settled, we will be able to reach certain conclusions regarding the

* Also see Chapters V and VI.

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

observable signs of this activity on cosmic scales and to analyze the
possibilities of detection of these signs with modern means.

The main factor which has been firmly and reliably established by
modern astrophysics is the universality of all the fundamental laws of
nature everywhere in the observable part of the Universe and over the
entire period of time covered by the evolutionary scale. We may there-
fore assume with fair likelihood that the physical laws known to us are
also known to supercivilizations. The knowledge of supercivilizations
clearly may cover a much wider gamut of physical laws, but the sum
total of their knowledge will contain as a subset all that we know. More-
over, the present level of our technical and scientific knowledge is
apparently an unavoidable and necessary step in the early development
of any technical civilization. We can thus try to formulate in crude terms
some general concepts applicable to all extraterrestrial civilizations.

A functional definition of a civilization is highly important for future use.
A detailed discussion of the functional definition of life, originally proposed
by Lyapunov /9/, is given in /1/ (pp. 125— 132):

a highly stable state of matter capable of developing
survival reactions using data coded by the states of the
individual molecules (D).

This definition adequately conveys the main content of the concept,
but in our opinion it has one fundamental shortcoming: it does not mention
the general laws and features governing the conception, development, and
evolution of various life forms. The life of any individual apparently can
be considered as a stochastic process governed by its interactions with
the environment and the state of the live object at any given time. The
evolution of the species in this case is regarded as a certain statistical
law which emerges from the growth and development of the individual
organisms. An obvious outcome of evolution is a steady accumulation
of information and its adaptation to practical applications. "Therefore,
it seems to us that the main statistical trend in the development of living
organisms is the tendency to gain the maximum quantity
of information about the environment and about the
organism itself (E).

For the lower life forms, this trend is dictated by natural selection.
This also seems to be the only stimulus for the development of the higher
forms of civilization.

The distinctive feature of the higher life forms is their ability to
undertake an abstract analysis of the acquired information. Systems of
living organisms begin to play an increasingly important role as the life
forms develop. However, we can hardly fix at this stage the exact number
of organisms and the structure of a high-level civilization. Thus, bypassing
the above definition of life, we can offer the following functional definition
of a high-level civilization:

a highly stable state of matter capable of acquisition,
abstract analysis, and application of information for the
purpose of extracting the maximum quantity of informa-
tion about the environment and itself and developing
survival reactions (F).

There is no need to include a specific coding mechanism in this general
definition. Information about environment and self includes all data about

EXTRA TERRESTRIAL CIVILIZA TIONS

animate and inanimate nature (including civilization), science, technology,
culture, art. (There are probably other, hitherto unimaginable fields
which also should be included in this category.)

If we accept definition F, the principal parameters characterizing
the degree and the character of development of a civilization are the
quantity of information and the rate of accumulation of new information
(e.g., the time to double the sum total of knowledge). Within the frame-
work of modern concepts (and here we have to differ with von Hoerner /2/,
p.278), it seems to us that definition F allows for an unlimited develop-
ment of civilizations. Von Hoerner's principal hypotheses regarding the
limit of development of civilizations include 1) total destruction of all
life, 2) destruction of intelligent life, 3) degeneration, 4) loss of interest.
These suicidal factors apparently acquire great significance for every
civilization at a certain stage of development, but there is no proof
that they are fundamentally unavoidable in every case for all
civilizations. The only reason for a civilization to stop developing in
the light of definition F is the existence of a finite quantity of information
in all the fields. This, however, seems to be a most unlikely propostion.

A highly important aspect for the search program is that the quantity
of information in certain fields is finite (this, naturally, does not imply
that the total quantity of information is finite). One of these fields with a
finite quantity of information is possibly space science at its present
level. To make this point, consider the following example. We have
already mentioned that the modern methods of astrophysics enable us to
study various objects in the Universe billions of light years distant from
the Sun. For these distances, the very concepts of length and time of
light propagation are not single-valued, and they significantly depend on
the particular model of the Universe used. The main method for estimating
the distances of extremely far objects is the determination of the change
in the wavelength of the emitted spectral lines (relative to the laboratory
wavelengths), i.e., the red shift 2. As we have mentioned before, spectra
of sources with z~ 2 have now been obtained. At the same time, radio
sources with the weakest observable continuous spectra may have a
substantially higher z. Were it not for absorption and scattering of
electromagnetic radiation in the intergalactic medium, the largest modern
radio telescopes could detect quasar-type objects with z~ 30, and the
projected radio telescopes could in principle advance this limit even
farther. However, calculations and statistical analysis of radio observations
show that this is not so.

The main factor preventing the effective observation of these ultra-
distant objects is apparently the scattering of electromagnetic radiation
by free electrons in the intergalactic and galactic medium. This effect,
as demonstrated in /10/, fixes z~5 as the most probable maximum dis-
tance at which radio sources are still observable (this is the value obtained
for a positive curvature model with qo = 1, H)= 300 km/sec . 10? light years,
and the present-day density of the intergalactic medium po 4 : 107% g /cm?),
Although no direct determinations of the density of the intergalactic medium
are possible at this stage, a statistical study of radio sources shows that
the number of weak sources is less than what could be expected without
Scattering. The theoretical result which points to the existence of the

1. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

maximum accessible distance thus appears to have a certain experi-
mental justification.*

The sphere characterized by maximum z contains a finite quantity of
matter, i.e., a finite number of cosmic objects. Since the structure of
celestial bodies is described by the same general laws in different parts
of the Universe, it is quite probable that the principal properties of
allthese objects will require only a finite time to study.

In all likelihood, many of the principal laws of nature will be
established within the next decade in view of the current tendency of
astrophysical research (A). Thus, the information concerned with space
Science has an objectively finite limit, and there is a definite possibility
that the supercivilizations may lose all interest in this science. This, in
particular, may resolve our dilemma (C) — there is no universal civiliza-
tion because the highly developed civilizations have lost all interest in
Space research. By space research we naturally mean research in the
modern astrophysical sense. There may be certain directions associated
with space science of which we are not aware at present (e.g., problems
connected with universal physical constants) and in which there is promise
of an unlimited quantity of information,

We should again stress that the problem of acquiring a complete
quantitative knowledge of the laws of the Universe is essentially simplified
by the inherent similarity of the celestial objects in various parts of the
Universe, as is evident from the currently available astronomical data.
Civilizations themselves are apparently the only type of objects which do
not follow this law of uniformity. Therefore, to ensure a maximum rate
of acquisition of new knowledge, the best way is to strive toward informa-
tion exchange between civilizations. In the light of modern ideas, exchange
of information through space is most effectively accomplished by means
of electromagnetic radiation. It is moreover clear that the most general
factor associated with the activity of supercivilizations is the use of mass
and energy on a gigantic scale.

In trying to distinguish between the activity of civilizations and the
effects of natural processes in the Universe, we should apparently be
guided also by the above definition of the civilization.

We cannot give any sound quantitative estimate of the maximum level
of development of supercivilizations. However, since there is a very
good chance of our mastering the entire electromagnetic spectrum and
thus markedly increasing the sum total of our astronomical knowledge,
we hope that this estimate will come within our reach some time in the
future.

The present-day astrophysical data do not impose any limit on the
possible development of supercivilizations, which in principle may reach
fantastically high levels. It may even be argued that the expansion of the
observable part of the Universe may conceivably be a result of some intel-
ligent activity of a supercivilization. According to the modern models of
the expanding Universe, all matter was in a superdense state some 10 billion
years ago. Does this preclude the continuous existence of civilizations at
earlier stages of evolution, 20, 100, and 1000 billion years ago, or is
there a possibility that they survived the instant when the Universe was

* Another reason which interferes with the observation of distant sources is the absorption of their radiation

by nearer sources. Already for z 2, the probability that the line of sight intercepts more than one object
is close to 1,

EXTRATERRESTRIAL CIVILIZA TIONS

in the superdense state? The age of the oldest civilizations can be reliably
fixed at a few billion years only when we shall have firmly established that
prior to the expansion the conditions in the Universe were adverse to the
inception and development of life.

Can we describe in general outline the development of a civilization
over cosmogonic periods? We know that many of the fundamental para-
meters characterizing the development of the Earth civilization grow
exponentially (see Chapter V). The time to double the scientific and
technical information is about 10 years, the time to double the power
resources, the raw material reserves, and the population is about
25 years. Extrapolation of the current rates of growth of our society
to the nearest future therefore leads to curious paradoxes.

In a book by a group of outstanding American authorities on thermo-
nuclear reactions /11/, the authors call our attention to the fact that
the quantity of energy that can be generated on the Earth is not very
high. There is a definite upper limit to it. The Earth absorbs (and
re-emits) 5-107 erg of solar radiation each second. To avoid drastic
changes in the Earth climate, the energy output of artificial installations
on the Earth must be limited approximately to one percent of this quantity.
Assuming a figure of 4. 10!? erg/sec for the current power output and an
annual growth of 4 percent, the authors show that the upper limit will be
reached in 125 years! This limit can be slightly stretched if we directly
harness the solar radiation. To this end, however, a considerable part
of the Earth's surface will have to be covered with solar energy con-
verters, a not very likely prospect.

Thermodynamic considerations show that this is indeed a fundamental
difficulty. After all, the entire expended energy is inevitably converted
into heat. And what then? Two solutions can be envisaged: either the
power output is maintained strictly constant after the allowed 125 years
of growth, or allthe forms of human activity involving large energy
requirements (industrial complexes and large-scale scientific experi-
ments) should be moved into outer space. The first alternative is
entirely unacceptable, since it virtually means that all further develop-
ment is stopped. The second alternative, on the other hand, appears
quite likely even at the present stage of development.

A similar conclusion regarding the inevitable expansion into outer
Space also emerges from an examination of other characteristics of
human activity (population explosion, chemical and radioactive con-
tamination of the ocean, insufficient open space on the Earth, exhaustion
of nuclear fuel resources, shrinkage of the biosphere, etc.). Power
difficulties, however, will probably prove the dominant motivating
factor. If a certain parameter P increases a factor of a annually, P,
will increase in t years to P=Poa', whence

t = ECT) years. (1.8)
The above estimate of 125 years was obtained using this relation. If the
growth rate a = 1.04 is maintained after the critical period, the human
power output will exceed the quantity of incident solar radiation after
240 years, after 800 years the total energy radiated by the Sun will be
exceeded, and after 1500 years we will exceed the total radiation output
of the entire Galaxy!

5780 26

I, ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

The population also grows exponentially, and possibly even faster, so
that there will be a steady pressure to maintain the exponential growth of
the other parameters. So far, our civilization has used up about 10" g
of mass. Assuming the same annual growth rate, the figure will reach
105! g in 2000 years, which is equivalent to the mass of more than ten
million galaxies! The quantity of information currently increases at a rate
of 10% annually; extrapolating for this rate of growth, we obtain an
increase by a factor of 109 in 2000 years, so that the quantity of infor-
mation by then will significantly exceed the total number of atoms in the
Universe (about 10%). Such a quantity of information in principle cannot
be stored or remembered! We thus reach the inevitable conclusion:
the current exponential growth constitutes a transient
phase in the development of the civilization and it will
be unavoidably restrained by natural factors.

Indeed, assuming a mean density of o for some space medium that the
civilization has set forth to harness, we see that, even advancing at the
velocity of light, it will be able to harness, after some time ¢, mass at a
rate not exceeding

M <n (ct)? ec, (1.9)
and energy at a rate not exceeding
E
LE « An (ct)? pc’. (1.10)

Hence it follows that the mass and energy requirements (and therefore the
growth of information, whose material carriers are mass and energy) may
increase exponentially only for a limited time, whereas an unlimited growth
may not be faster than . For our civilization, as we have seen, the dura-
tion of the future exponential growth phase can be estimated at about

1000 years. And what then? Since the development of power resources

on the Earth (and, in general, in any finite volume) is limited by thermo-
dynamic considerations (the overheating effect that we have mentioned
before), future economic growth after some 100— 200 years will probably
push humanity into outer space! This, in our opinion, is the objective
tendency and the main task of space exploration at this stage.

Should not the nonexponential growth be interpreted as a sign of a
decaying civilization? In our opinion, even a linear growth of information
indicates a viable civilization, Indeed, a constant rate of acquisition of
information signifies that a constant quantity of new, highly significant
and highly valuable data is acquired every year. "This in no way obstructs
the main tendencies in the development of civilizations. The so-called
"feedback effect" will apparently constitute a decisive factor for further
development of our civilization. Everything depends on whether
supercivilizations exist or not, If the answer is in the affirmative, recep-
tion and assimilation of information from supercivilizations may play
a leading role in future development. This learning stage may lead to a
rapid jump of the civilization to the highest level. 1f we assume that
every civilization at a certain stage of its development passes through
such a learning stage, we conclude that there will be virtually no civiliza-
tions in an intermediate stage of development or in a stage close to ours.

EXTRATERRESTRIAL CIVILIZA TIONS

The second possibility — total absence of supercivilizations — will
apparently necessitate a complete revision of our current ideas of unlimited

growth and development.

$5. THE SEARCH FOR SIGNS OF ACTIVITY
OF SUPERCIVILIZATIONS

The general considerations of the previous sections lead to certain
conclusions regarding the types of activity of supercivilizations which can
be detected at the present level of development.

The most general parameters of this activity are apparently ultra-
powerful energy sources, harnessing of enormous solid masses, and
transmission of large quantities of information of different kinds through
Space. In this section we will consider the first two parameters which
are a prerequisite for any activity of a supercivilization.

Energy sources

As we have noted before, the present-day astrophysical observations
do not provide any indication of the existence of an upper limit for the
energy output of a supercivilization, This limit, however, will probably
emerge when we have covered the entire electromagnetic spectrum, from
108 to 10!? Hz. This interesting conclusion follows from basic thermo-
dynamic considerations: the entire energy expended by a supercivilization
is inevitably converted to heat. This thermal energy cannot accumulate
indefinitely inside a closed volume, to avoid critical overheating. The
only way in which this heat can be dissipated is by radiation into outer
Space. Any power system thus inevitably involves eventual radiation of
its entire power output in the form of heat into space. If the efficiency
of these systems is very high, the spectrum and the surface brightness
of the radiating body should correspond to the blackbody spectrum at a
temperature equal to the effective temperature of all the forms of electro-
magnetic radiation received from outer space (the equilibrium temperature
in the intergalactic medium is around 3°K). It is quite probable, however,
that the efficiency of these power systems is less than 100% (there can be
various operational reasons for this). The resulting emission spectrum
is more complex. It is difficult to predict the specific features of sources
of this kind. The only reasonable thing to do at this stage is to concentrate
on radiation sources with maximum bolometric power. Quasars are the
only known objects which fall under this category.

Let us briefly describe the main regular features established for these
remarkable objects /8/.

The radio emission of quasars was discovered more than 10 years ago,
but the widespread interest in these objects was aroused only recently.

In 1960— 1962, following a substantial improvement in the directivity on
radio observations, it was established that some radio sources have the
Same coordinates as star-like optical objects. Prior to that time, the
consensus of opinion had been that most radio sources are identifiable

I, ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

with large galaxies. It thus appeared that a new class of stars with
anomalously powerful radio emission had been discovered in the Galaxy.
Further observations, however, proved this hypothesis to be wrong.
The observational data indicated that these were an entirely new type

of extragalactic object, of which nothing had been known before. (Note
that this again stresses the need to cover the entire electromagnetic
spectrum in our observations.)

When observed through optical telescopes, the quasars appear as
star-like objects in the sense that the apparent angular diameter is
substantially less than the resolution limit of the astronomical optics
(fractions of an angular second). Near some of the quasars, nebulous
filaments are observed, which may be irregular in shape or follow a
general radial direction from the star, reminiscent of ejected gases.
Figure 4 is a photograph of one of the nearest and brightest quasars,
the radio source 3C 273, with a noticeable ejection on top right. The
ejection is no wider than 1"— 2", it begins at a distance of 11" from the
star and terminates at a distance of 20". Ejections and filaments are
also observed near the quasars 3C 48, 3C 196, and 3C 279.

FIGURE 4. The quasar 3C 273.

One of the most remarkable features of the optical spectra of quasars
is the exceptionally strong red shift of the spectral lines. The red shift
varies from 0.158 (for the nearest quasars 3C 273) to values corresponding
to a three-fold change in wavelength (3C 9, see Figure 2). This unusually
high red shift, if interpreted as the result of the expansion of the Universe,
points to tremendous distances and fantastic luminosities of these objects.

Because of the high red shift, the optical spectrum contains some lines
which normally lie in the ultraviolet in laboratory spectra. For ordinary
stars and galaxies, this spectral region is inaccessible to observations
from the Earth, as the atmosphere is opaque to wavelengths shorter

EXTRATERRESTRIAL CIVILIZATIONS

than 30004. The spectra of quasars have by now been studied down to
1000 À, and some spectra actually gave the profile of the L, hydrogen
line — the strongest line of most cosmic objects. Table 1.1 lists the
elements and the ionization stages discovered in the spectra of quasars.
The first column gives the elements in the order of increasing atomic
number, the second column itemizes the observed ionization stages.
The missing lines are apparently those of elements which occur in small
quantities, in accordance with their normal abundance, or which nor-
mally do not have bright lines in the observed part of the spectrum.

The logarithm of the normal abundance (by number of atoms) is given
in the last column of the table.

TABLE 1.1
Element Ionization Abundance Element Ionization Abundance

H I 12.0 P = 5.53
He I 11.16 8 I] 7.22
Li = 3.0 cl =a 5.4

Be > 2.4 Ar IV 6. 62
B = 2.8 K = 4. 88
C II, I, IV 8.48 Ca JI 6. 22
N IV, V 7.96 Sc = 2. 91
o 1 IL III 8.83 Ti ni 4.82
F = 5.4 V PY 3. 78
Ne IH, V 8.44 Cr IH 5. 38
Na = 6.22 Mn I, HII 5.10
Mg IL V 7.46 Fe II 6. 90
Al IL HI 6.28 Co I 4.72
Si I1, III, IV 7.41 Ni II 5. 93

The conditions of excitation of spectral lines in quasars are apparently
highly variable. Some quasars show mainly emission lines, most of which
can be identified with the spectra of certain elements. Figure 5is a
microphotometric tracing of the spectrum of 3C 273 /12/. In addition
to emission lines, the spectrum shows wide emission bands of uncertain
origin. Figure 6 is the profile of the H, line in the spectrum of this
quasar /13/. Some features of the line profile show distinct signs of
a shift relative to the line center. The Doppler velocities corresponding
to this shift are as high as a few thousands of kilometers per second.
Some quasars have a rich spectrum which also contains absorption
lines and bands (e.g., 3C 191). Some quasars (e.g., 3C 682) do not
show any lines at all.

The continuous optical spectrum of quasars also shows a number
of characteristic features. The energy distribution in the quasar
spectra is markedly different from the energy distribution in the
stellar spectra. Quasars can thus be readily identified in large-scale
measurements of star color with light filters. The energy distribu-
tion in the optical spectra of quasars is adequately described by a
power function F,cv, and a probable mechanism is therefore
emission or scattering of radiation by relativistic electrons.

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

| 9969-(4024)

AUAM

]/Ne I] 3868-14485)
] n 3970-4603)

] Hy 4702-14756)
4179-4846)

Ed
z

) Hy 9940/5030)

| 4997-(8263) Wy

Hg 4861-(5637)
[OZ 5007-(5606)

] 9254-16095)

| 5572-16464)

ae tone

LMA Ferree À-6 -4 -2 0 +2 +4 +6100 kmlsec
FIGURE 5. The microphotometric trac- FIGURE 6. The profile of the Hg line of the quasar
ing of the spectrum of the quasar 3C 273. 3C 273.

Figure 7 is a plot of the optical colors obtained with three filters,

U (à 3600 Å), B(x 4400A), and V(A5500À). The horizontal axis gives
the B — V difference, and the vertical
axis the U — B difference for the same

pud 7 T object. The lower curve is the locus
-72 corresponding to the main- sequence
stars, and the top line is the power
-488 q energy spectrum. The quasar
me region is cross hatched.
f One of the most puzzling pro-
Z2 perties of quasars are the variations

S
SN of their intensity. Prior to the dis-

24 covery of quasars, extragalactic
08 astronomy was generally assumed
“04 -Q2 Q 402 Q4 46 48 p fe to deal with highly stable sources.
The brightness of galaxies remains
FIGURE 7. The colors of main-sequence stars constant over billions of years
and quasars. (except for the brief supernova

EXTRA TERRESTRIAL CIVILIZA TIONS

explosions). And yet, the observations of the first quasars have shown
that their luminosity is significantly variable. Using old photographs

of the sky, the astronomers managed to reconstruct the light curves of
these objects over a relatively long period. Figure 8 shows the smoothed
light curve of 3C 273 for the period 1888— 1963 /14/. The mean light
variation period of this source is about 9 years. The mean photographic
magnitude of 3C 273 decreases according to the equation

mg, = 127.47 + 37.67 (T — 1900),
+0.08 047

(where T is the year of observation), which gives 300 years for an
exponential decrease of brightness to 1/ e /15/. Faster brightness fluctua-
tions, whose statistical character is still unclear, have also been observed.

Z, years

!
7900 MES SDN

FIGURE 8. The smoothed light curve of the

quasar 3C 273 corrected for the secular decrease
in brightness.

Figure 9 plots the results of photographic and photoelectric measurements
of the stellar magnitude with a B filter for the quasar 3C 446 /16/.
Occasionally, the brightness of this object changes by as much as a
factor of 2in 24 hr! This rapid variation of brightness provides a
direct estimate of the size of the emitting region — less than one light
day (< 3- 10? cm), i.e., much less than the size of a galaxy (tens of
thousands of light years) and probably even less than the size of the
solar system: the diameter of the orbit of Pluto is 0.5 of a light day.

Both the continuous and the line spectrum of quasars apparently
change (the changes cover line widths, line intensities, and wavelengths
/17/). The correlation between these variations has been hardly
studied.

Some quasars show a considerable linear polarization of the optical
radiation. The same quasar 3C 446 has a maximum difference of 0",2
between the intensity of the perpendicular polarization components.

The degree of polarization and the position angle apparently change with
time. The polarization of the infrared radiation at à= 1.64 for the
quasar 3C 273 reaches 40%.

Let us now consider some fundamental results of radio observations
of quasars. The angular resolution of the modern radio telescopes can
be made as high as 0".001 (by using interferometric techniques, observing
the diffraction pattern during lunar occultation of radio sources, and

I, ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

studying the radio flux fluctuations associated with radio wave propagation
inaninhomogeneous interplanetary medium), and this is considerably higher
than the resolution attainable with optical telescopes for the same objects.

-2cm

Flux at À

v
i-i
B
&
Eo
«
B
=
ad
g
F3
[s)
Eum 9] m dozd A a)
ese ke weer VAS §F Sg Hs &
o a 3 [7] a, 8
S2Ssee 25 2cR52A8L SSE B

FIGURE 9. Light curves of the quasar 3C 446.

However, not much information has been obtained so far by the new radio
methods. The observations of 3C 273 (the best studied quasar) revealed
the existence of two sources: source A corresponding to a luminous ejec-
tion on the photograph of this object, and source B which adequately
coincides in position with the quasar itself. Figure 10 shows the radio
spectra of components A and B, which are markedly different /18/.

enn vis
Ww
mHz
E M
0 de
i!
\
\
M
Halo \
p id
g” $03 913401007 pee
p 7° m3 107 70°

FIGURE 10. The spectrum of the components of 3C 273.

EXTRA TERRESTRIAL CIVILIZA TIONS

Source A is elongated along the optical ejection and grows brighter at the
outermost end, where its angular dimensions are 5" X 1".5. Source B in
its turn consists of a spherical halo some 6" in diameter and a central
nucleus /19/. Radio-interferometric observations reveal that most of
the energy is radiated from a region not exceeding 0".002 /20/.

The spectra of quasars often deviate from the normal power function,
and this probably suggests a complex structure or a variety of emission
processes. The most interesting properties are those of sources with
peculiar features in the short- wave part of the radio spectrum. Figure 11
shows the spectrum of 3C 279; like the spectrum of 3C 273B, the
radio flux shows a tendency to increase toward shorter wavelengths.

a 100 AU 00 AU 10000 20000
MHz

FIGURE 11. The spectrum of 3C 279.

The radio emission of these objects is generally variable. Figures 12
and 13 plot the time variation of the radio flux from these two sources
at various wavelengths /21/. Particularly strong and rapid variations
are observed in the millimeter range. In 1966, a decision was taken

to launch an international program of systematic observations of selected
objects in the entire electromagnetic spectrum in order to study the
variability of quasars. The sources 3C 273, 3C 279, 3C 345, CTA- 102,
and others were chosen for this purpose. The list also included the
source 3C 84, which is a nucleus of the anomalous galaxy NGC 1275.
The properties of this source have much in common with the properties
of quasars. Detailed observations also reveal a deep- running analogy.

No individual radio lines from quasars have been observed thus
far, since every quasar requires special receiving equipment adjusted
to its red shift.

The brief description of the observational data shows that our infor-
mation about quasars is highly deficient even in the well- mastered
frequency ranges and for the brightest sources. It is quite probable
that some of these sources have an exceptionally strong radiation in
the intermediate spectral region (between the radio and the optical spectra).

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

Flux /2 W/m? uz

Flux: W/m? Hz
42 T ; a 22
54 ep C279
£0 18 AZ
52 74 S
48 pA x
44 No
40 `-
IG 16 | o

32
28
24

6
aa 7H Ns
40 E14 2
536 5 m
a2
$4, . 7

S S S
3N
wrer moer roor

HHL
SiS
is a aai
|

46 22 72
42 m
I8 8

+g o 40

1 A L LI. 1
7962 1964 7955 1962 1964 1966
FIGURE 12. Variation of radio flux, degree of polariza- FIGURE 13. Variation of the radio flux from the
tion, and position angle for the quasar 3C 273 at quasar 3C 279 at 8000 MHz.

8000 MHz.

The bulk of the energy of 3C 273, for instance, is definitely known to be
radiated in this range. Figure 14 shows the combined spectrum of 3C 273B
based on both radio and optical observations, plus the new measurements
in the millimeter and the infrared spectra /23/. The steeper short wave
curve is based on the 1964 measurements, and the gentler curve is the
result of 1966 measurements. The spectrum of quasars probably

extends far into the ultraviolet and the X-ray region. Recently, 3C 273

was apparently found to emit at 1— 10 À /24/.

The total bolometric luminosity of the quasars is unusually high. The
total flux emitted in the infrared and in the submillimeter region by 3C 273B
reaches 4.107? W/m?. Since the distance to the source is < 1.5 - 107 cm,
the total energy radiated in this range is about 1047 erg/sec. The energy
emitted in the optical spectrum is !/j of this value, and that in the radio
spectrum 499 of this value. Thus, there apparently exist quasars which
are basically infrared sources /25/. The bolometric power of quasar 3C 273
is thousands of times greater than the corresponding power of the giant
galaxies.

Studies of the spatial distribution of quasars revealed still another
remarkable feature: quasars never occur in clusters or near individual
galaxies /26/.

EXTRA TERRESTRIAL CIVILIZA TIONS

mz 302738

5 7 8 9 WW 12 13 u i5 IB
1g vHz

FIGURE 14. The spectrum of 3C 273B according to radio,
infrared, and optical measurements.

Very interesting conclusions emerge from statistical studies of the
distribution of quasars according to the observed radiation flux. These
statistics reflect the line-of-sight distribution of sources. If we allow

for the time of propagation of the radia-
tion, this distribution can be taken as
N ste! characterizing the quasar number and
power at various stages of evolution of

m*
the Universe. Figure 15 plots the func-
tion N(F,), i.e., the number of all radio
m^ sources brighter than a given flux F, vs.

the flux. The curve is based on the
observations of all the radio sources
v? at 178 MHz up to a maximum flux of
5-108 W/m?.Hz /26/.
Theoretically, a uniform distribution
of radio sources in a Euclidean space
without expansion gives N(Fy)o Fy".
D x 7 Z y It follows from the theory that the
bie ae red shift associated with the expansion
of the Universe should lead to a more
gentle dependence. Observations, on

FIGURE 15. The number of sources
the other hand, give a steeper dependence:

brighter than a given flux vs. the flux
t 178 MHz. -
value à Z. N(F) a FS.

Recently it has been established that
if all the sources are divided into two groups — quasars and radio galaxies —
each class will have its own distribution function N(F,). For radio galaxies
N(Fy) « Fy", and for quasars N(Fy)o F3? /27/.
The possible reason for this steep distribution is the rapid evolution of
the radio sources in an expanding Universe (either a decrease in the number
of sources in every bounded volume with the expansion of that volume, or

I, ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

a decrease in the brightness of the source, or both). There is a possibility
that only quasars are characterized by this exceptionally fast evolution.

If the division of sources into two groups is justified, the number of
quasars among sources with fluxes of 10779 W/m?.Hz is comparable with
the number of other sources (from observations at 178 MHz). The curve
N(F,) then also shows a marked saturation at low fluxes. The low bright-
ness temperature of the extragalactic background (about 20°K at the same
frequency /26/) also points to the existence of a certain limit number of
sources. From these results, we can find the time at which the forma-
tion of quasars began. This was approximately one billion years after
the Universe began expanding.

The exact nature of quasars is unknown at this stage and is very diffi-
cult to.guess at. The discovery of the variable radio flux rendered all
the conventional mechanisms of radio emission inadequate. The fairly
rapid fluctuations of the radio flux are difficult to reconciliate with the
emission mechanism of relativistic electrons moving in magnetic fields.
Another alternative is to invoke coherent emission mechanisms (e.g.,
plasma oscillations /28/ or coherent stimulated emission of relativistic
electrons /29/).

What is the probable structure of a quasar according to current notions?
The core of a quasar is a nucleus measuring <10' cm, whose mass is
approximately 10° solar masses. The nucleus plays a definite role in the
overall behavior of the quasar. In particular, its emission constitutes
the main contribution to the continuous spectrum of the source. The
nucleus is a giant star where the equilibrium is maintained by a balance
between the gravitational energy and the energy of magnetic turbulent
plasma or the rotational energy of the spinning star. The energy losses
through the powerful radiation of the nucleus are made up by the gradual
contraction of the star, i.e., by the gravitational energy resources. It
follows from the theory of gravitational collapse that when a mass contracts

to its gravitational radius rea 7, it releases energy which amounts to

several tens of percent of Mc? (thermonuclear reactions release only about
0.5% of M? ). ForM=108Mo, rg 7 3.10 cm, i.e., a figure of the order
of magnitude of the diameter of the Earth's orbit. The energy resources
corresponding to (5) Mc? are equal to 6. 109! erg, which is sufficient to
keep a quasar going for 20 million years at a rate of 1047 erg/sec. The
existing estimates of quasar masses, however, are highly uncertain, and
they probably provide only a lower limit (the mass of the nucleus is taken
to be larger than the mass of the surrounding envelopes, which can be
determined from emission and absorption lines). The activity of the
nucleus is associated either with its pulsations or with the fact that it
constitutes a close binary system of high-mass superstars. This activity
involves ejection of ionized gas and streams of relativistic particles.

It is quite probable that at the center of galaxies, and in particular at

the center of our Galaxy, quasar-like objects exist. A certain region

at the center of our Galaxy emits strong nonthermal radio emission.

The motion of ionized and neutral gas clouds in the central parts of
galaxies is also reminiscent of quasars. A nucleus with bright emission
lines was discovered at the center of the Andromeda Nebula (M 31).

We have mentioned before the striking similarity in the optical and

EXTRA TERRESTRIAL CIVILIZA TIONS

radio spectra of the nucleus of the galaxy NGC 1275 and of quasars.
However, effects of this kind in galaxies are many orders of magnitude
less powerful than the corresponding phenomena in quasars.

It should be noted that the nature of many of the known radio sources
is no less puzzling than the nature of quasars. For example, some
double galaxies, including one of the brightest radio sources in the sky,
Cygnus A, are great cosmic enigmas. The optical nebula located between
the two radio sources is not a galaxy in the usual sense of the word. It
seems to be made up entirely of high-temperature gas. Recently the radio
galaxy has been shown to emit high amounts of energy in the X-ray spectrum.
The radio galaxy Virgo A emits in the X-ray spectrum 100 times as much
as in the radio and the optical spectrum /30/.

+80" +60" +40" +20" O -20° M -60" -80°

FIGURE 16. The structure of the radio galaxy Cygnus A
at 11 cm.

Figure 16 is a chart of Cygnus A obtained at 4 — 11 cm with a radio
interferometer. The structure of this object is clearly very complex,
and it contains several sources of small angular dimensions. Figure 17
is a photograph of the sky near the double radio source 3C 33 /31/.

The radio source components show on the photograph as two ellipses
which give an idea of the source size and correspond to a certain peri-
pheral enhancement. Midway between the sources we see a galaxy,
which apparently brought forth the two objects. Some of the puzzling
questions are what caused this "ejection" from the galaxy, how to
explain the striking likeness in the radio spectra of the ejected sources,
what prevents this formation from expanding through the interstellar
medium if these are indeed relativistic gas clouds, as many seem to
think?

The most remarkable objects of this kind are the radio sources 3C 343
and 3C 343.1 /32/. Their spectra are also perfectly identical, the distance
between the components is 29', the angular size of each component is less
than 0".1. The parent galaxy has not been discovered so far. The identi-
cal spectra of two complex cosmic objects whose separation from one
another is greater than the diameter of a sizeable galaxy are very diffi-
cult to account for by any of the known natural mechanisms.

Let us summarize. It is obviously too early to suggest that
quasars or some of the radio galaxies are artificial sources of energy.

1. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

FIGURE 17. The area around the radio source 3C 33.

It seems, however, that this hypothesis definitely deserves more than a
cursory glance. Anyway, this hypothesis has stimulated during recent
years some discoveries of highly important properties of quasars, which
are discussed in the next section. Observations have thus far established
that quasars are the most powerful and yet the most compact energy sources
among all known astrophysical objects (the quasar nucleus is smaller than
the solar system whereas its radiation is more powerful than that of a
thousand galaxies!). Future surveys in unmastered frequency ranges will
show whether or not more powerful sources exist in the Universe. Studies
of the most powerful objects will clearly enable us to fix an upper limit

to the permissible energy output of a civilization.

Solid matter

From the point of view of modern physical concepts, the only state of
aggregation of matter which is capable of storing indefinitely a large
quantity of information is the solid state. The main feature of the solid
State is the fixed and constant arrangement of atoms in the lattice. This
feature is the basis of modern technology, in that it ensures constant and
immutable properties of constructions; the same phenomenon made
possible the development of biological processes on the Earth.

The solid matter also probably provides the basic constituent for the
technology of supercivilizations, in particular in various data acquisition
and processing systems. Therefore, a discovery of solid cosmic objects
may have a significant bearing on the solution of our problem.

EXTRA TERRESTRIAL CIVILIZATIONS

Unfortunately, the solid state of matter is the most difficult to detect
in the Universe, because of its low temperature and the correspondingly
weak emission of radiation. Therefore, our information on the quantity
and properties of solid matter in the Universe is virtually nil.

The astrophysical data in our possession refer to planets and their
satellites, to meteorites and interplanetary dust in the solar system
(also measured from rockets), and to the extinction of stellar light by
interstellar dust particles and its scattering in the reflecting nebulae.
The properties of interstellar dust are mainly derived from theoretical
considerations regarding the quantity of the heavy elements and the
possible properties which cause mechanical destruction of the dust
particles, their heating and cooling.

High-mass solid objects in the Universe are extremely intractable.
Let us consider this point in some detail. There can be two different
approaches to the search for these large solid objects: trying to detect
the nearest individual massive objects and trying to detect the combined
emission (or absorption) effect of a large assembly of solid bodies. Let
d, 6, and T be respectively the size, the density, and the surface tem-
perature of the solid objects, n the number of these objects in unit volume,
l the size of that part of the Universe which is filled with these objects.
The mean density of matter associated with the solid objects is then

p =n ôd?,
and the angular size of the nearest object is

Pmax =dn = (£)" , ( 1.1 1)

the observed emission temperature of a large assembly of such objects
(treated as the background emission) is

Tg = Tnd", (1.12)

and the optical thickness for light absorption or scattering by these solid
objects is
t=nd". (1.13)

Assuming that the concentration of solid matter with ô~ 1 g/cm? does
not exceed the mean density in the Universe p —10 7? g/cm? (for extra-
galactic solid objects) or the density in the Galaxy for galactic objects
(either estimate is grossly exaggerated), we find that the angular size
of the nearest objects does not exceed 4- 10? and 4-10? sec, respectively.
Since the surface temperature of these objects is limited, there is no way
to detect them individually.

The combined emission of a large assembly of solid objects will be
difficult to observe when T4 xT., where T, — 3°K is the equilibrium tem-
perature for all types of electromagnetic radiation in the Universe.
Absorption effects are difficult to distinguish when t<1. Since the
surface temperature of the solid objects clearly should be greater than
(or equal to) T.~3°K, the two conditions combined give the inequality

L ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS
nd?!l«1, and since nase, we find
d> $; (1.14)

assuming that the solid matter accounts for a noticeable fraction of the
density in the Universe p —1077? g/cm? (this estimate is a gross exaggera-
tion), and taking for the density of the solid objects 6 — 1 g/cm? and

I~ th ~ 1028 cm, we find that even in these extreme conditions particles
measuring d>1 mm remain absolutely undetectable.

In our opinion, this difficulty is virtually insurmountable in the sense
that giant solid constructions of supercivilizations may remain undetectable
even with the largest telescopes. The attempts to detect solid objects
from their gravitation effects are also absolutely hopeless, since the
existing estimates of the total mass of star clusters, galaxies, and clusters
of galaxies are characterized by low accuracy (mainly because of the high
proportion of low-luminosity stars). The estimates become more encouraging
if we assume that the effective density of constructions in the Universe is
6< 1. One of the examples of constructions of this kind is Dyson's sphere,
a shell enclosing a star, with a radius of about 1 astronomical unit. The
equivalent density of this construction is

2
Seq ~ p SA _ 36h

ToO

yee
where r~ 1.5. 10? cm is the radius of the sphere, A —10? cm is the thick-
ness of the sphere, and à— 1 g/cm?, In this case, we find ðe ~ 2 - 10 !! g/cm?,

0?9

the mass of the sphere is M =F arde ~3-1 g (approximately half the

mass of the planet Saturn), and the visible angular dimensions (1.11) for the
Metagalaxy and the Galaxy, respectively, are <0".15 and <15". These
objects can be detected with modern telescopes. More detailed cal-
culations /33,34/ lead to the estimates listed in Table 1.2.

TABLE 1.2

In the calculations of Table 1.2 it was assumed that the heat emission
of the sphere (with a surface temperature of 300°K) at wavelengths between
8 and 13 microns was observed with modern high-sensitivity bolometers
and optical telescopes. In this table, P is the bolometer sensitivity, Dis
the telescope diameter, and R is the maximum distance at which the
thermal emission of Dyson's sphere is detectable, assuming a minimum
signal/noise ratio of 9. Note that these observations can be easily carried

EXTRA TERRESTRIAL CIVILIZATIONS

out with the existing telescopes on Earth, since the 8—13y range
corresponds to one of the transparency windows of the Earth's atmosphere.
In general, the thermal emission peak of solid objects at temperatures
between 3 and 300°K falls between 104 and 1 mm, and at these wavelengths
the atmosphere is highly opaque, mainly due to the absorption by water
vapor. The search for these sources should therefore lean heavily on
Observations from beyond the atmosphere.

$6. THE SEARCH FOR INFORMATION TRANSMISSIONS

In the previous section we discussed the search for the various signs
of activity of civilizations. One of the most probable elements of this
activity is apparently transmission and exchange of information. These
transmissions can be divided into two broad types: 1) exchange of infor-
mation between highly developed civilizations of approximately the same
level, and 2) transmission of information aimed at raising the level of
less developed civilizations. 1f supercivilizations actually exist, trans-
missions of the first group may prove virtually inaccessible to us (e.g.,
these transmissions may be directed by tight-beam systems and the
transmission line need not necessarily intercept the solar system). On
the other hand, transmissions of the second group, by their very nature,
should be readily accessible and easily detectable by others. The
reception of transmissions of this kind is expected to have a funda-
mentally significant influence on the development of our civilization
(von Hoerner's feedback effect /2/), and as a result we will rapidly
rise to the highest level of civilization currently existing in the Universe.
Probably the fastest and the simplest (though the most fantastically
sounding) way to achieve this advancement is by merging with the
nearest supercivilization.

How is the search for transmissions of this kind to be planned?

Redundancy considerations indicate that we can hardly expect a great
number of transmissions of this kind. We will do better to concentrate
on one or several sources of electromagnetic radiation which stand out
among the rest in terms of their intensity or some other property. The
search for these prominent sources can be effected by means of sky
surveys in the least noisy frequency range. As we have noted above,
the technical means for the detection of electromagnetic radiation have
improved to such an extent that instrumental noise is no longer the main
limiting factor. In the next 5— 10 years, apparently, the receivers will
attain their maximum sensitivity for astrophysical work, which is deter-
mined in each frequency range by the intensity of background radiation
and by random fluctuations of the signal. Figure 18 plots the background
intensity spectrum for an observer situated in the intergalactic space, far
from the bright galaxies. This spectrum has been reconstructed from
the results of measurements in the radio, optical, and partly X-ray
Spectra, and also between the optical and the X-ray spectra the curve
is based on theoretical calculations, which take into consideration the
emission of the interstellar dust in galaxies and the total emission
of the galactic stars, and also on extrapolation of the available results

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

of observations /35,36/. For an Earth-bound observer, the background
radiation of our Galaxy has to be added to this spectrum. The resultant
background intensity from the Earth is shown in Figure 19 for the "brightest"

(the center of the Galaxy) and the "coldest" (the Galactic pole) parts of the
sky.

I
3
omat”

b dede bod

ta dd. donc cp qo p Tear
678 9101) 12 19 1415 1617 18:19 2021 2222 Mv iz

FIGURE 18. The spectrum of the background electromagnetic
radiation for an observer in the intergalactic space.

—————————— —— cR

N,
4 hy

' roto At Lo dad ra
6 7 g ZU 1213 1415 18171818 BT BEB
v

FIGURE 19. The spectrum of the background electromagnetic
radiation for an observer in the solar system.

Both spectra show deep intensity minima, and these valleys are
apparently the most suitable for interstellar communication, The discrete
(quantized) nature of the electromagnetic radiation is another significant

EXTRATERRESTRIAL CIVILIZA TIONS

factor to be considered in connection with the choice of the transmission
range. The distinctive feature of the spectra in Figures 18 and 19 is that
the background intensity is everywhere higher than the blackbody intensity
at 3°K. The range with the minimum equivalent blackbody temperature
(the region where the "relic" background radiation predominates) lies

at wavelengths between 3 m and 30 cm in Figure 18 (this range is some-
what narrower for the case in Figure 19). In both figures, the dashed

line marks the limit where for a blackbody radiation ae = 1, and con-

2hv* 1 2hv 1

Fw" F gl
ett -1

have 4% *T 771 and the quantum effects are therefore most prominent.

To the right of this limit we

sequently /,=B,=

The solution of the problem of optimum signal transmission against
a noisy background essentially depends on the particular parameters that
are to be optimized. Allowance for quantum and classical fluctuations
leads to the following expression for the maximum quantity of information
which can be received in unit time in a unit frequency interval /37/:

Av
2m a [Pog E NS URL ANTT
ey=tn[t + i-e x)| + hv + nn PT RE 1 Av . (1.15)
hv e*T —1 hv t .hv QE -1
eT =]

Here P, is the power spectrum of the received signal at the receiver input,
T=T(v) is the effective temperature of all the noises corresponding to an

approximate input noise spectrum of the form e,= E . For fixed P, and

eT |
T, the function c, decreases monotonically with increasing frequency
(Figure 20).

I
7
a

27,
DEED A EE UE

FIGURE 20. Cy vs. frequency.

In the classical case m 1), equation (1.15) takes the form

E neo 19

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

If we retain only the first term of the expansion, (1.16) coincides with
Shannon's standard expression for the rate of information transmission. *

In another limiting case Dc» 1, and assuming that the signal is much

more powerful than the noise ee m: i we get

eTa]
= Py) 4 Py Ave
e, 2 ln (1 A) +4 In (14 2). (1.17)
i.e., the rate of information transmission depends only on the number of
quanta of the signal received in unit time >. For high spectral intensities
of the signal RR» i), equation (1.17) takes the form
cy=In(1+)41, (1.18)

which is also close to Shannon's expression if we introduce the "equivalent
temperature" of the quantum noise kTeq =hv.

For an ideal detector, the noise s,is determined by the intensity of the
Sky background radiation /,, i.e.,

ely
2y? *

(1.19)

1
&, = p LRA, =

Here A, and Q, is the effective collecting surface and the effective solid
angle of the receiver at frequency v (the antenna, if radio frequencies
are considered), and the factor 4 allows for one polarization component
of the intensity /, (both components are assumed to have the same intensity).
The last part of equality (1.19) is valid only if 4,Q,=42, which is not always
true. For example, for the optical telescopes, the minimum solid angle
Q, is general!y determined by the scattering in atmospheric inhomogeneities,
and not by the diffraction pattern of the point source. As a result, the
angular size of the source is very seldom less than 1", and the adjustable
aperture used to restrict the background should not be less than this
figure. Thus, we always have A,Q, > 22 and the equality in this relation
(corresponding to a pure diffraction image) ensures the best signal/noise
ratio.

The signal power spectrum at the detector input is related to the
radiation flux of a point source F, by the equality

P= RAF, (1.20)

tol -—

where the factor Yz allows for the polarization components, as in (1.19).
These general relations make it possible to assess the peculiar features
of signals of artificial origin.
The reception of signals from extraterrestrial civilizations can be
divided into three stages: 1) search for call signals and their decoding,
2) search for the key to transmission and its decoding, 3) reception
and decoding of information.
Let us consider in some detail the first phase of the procedure, namely
the search for call signals and the choice of the most suitable frequency range.

* See also Chapter HI.

EXTRA TERRESTRIAL CIVILIZA TIONS

Call signals are intended to facilitate the detection of the source, and
they carry a certain minimum quantity of information which is sufficient
to firmly identify the source as an artificial object.

The choice of the optimum frequency range for call signals thus amounts
to the following. We have to'find the frequency v and the operating condi-
tions of the receiver which ensure the maximum signal/ncise ratio for a
given total energy flux from the source per unit surface area near the
Earth F and the given search time fy.

If an ideal receiver is used, the root-mean-square noise power at the
input is determined by the fluctuations associated with the natural back-
ground radiation from outer space. Allowing for the fluctuation in the
number of photons, we have

V AP? = [e + ev] eae (1.21)

where e, is defined by (1.19), and Avand t are the receiver band width and
time constant, respectively. Seeing that the input signal power is P,=

=P, Avi FA,, we find for the signal/noise ratio

1
is zf^Y* (1.22)
[e + ehy] ‘hV dv C

The entire search time tf incorporates both the frequency search and
the direction search, so that we have to maximize N for a given to,

dyes ite (1.23)

As we have noted before, the real reception conditions are such that
the solid angle Q, of the receiver and the effective collecting area A, are
related by the equality

QA, = RM, (1.24)

where the numerical coefficient £,21. Because of the great difficulties
in the manufacture of large precision surfaces, the coefficient k, increases
with the increase in frequency. The actual conditions of propagation of
light and radio waves in the atmosphere also increase the coefficient k,,
and this effect is particularly pronounced for observations in the optical
region. In observations from outside the atmosphere, allowance should
be made for the increase of angular dimensions due to the propagation
of radio waves in the interstellar and the intergalactic plasma. This
problem is discussed in detail in Chapter II, where it is shown that the
scattering is negligible in the centimeter and the decimeter range, but
it may reach significant values for the meter wavelengths.

Using (1.19), (1.23), and (1.24), we write (1.22) in the form

N= F V Avvto (1.25)

Se fe OR
Vin c V De, 1, H

1. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

The following conclusions can be drawn from this relation. First, to
obtain the maximum N, we should make &,as small as possible, if the

main contribution comes from classical fluctuations US > Z, This

condition is always imposed in the region where the background is described
by an effective temperature which satisfies the inequality rs «1. However,
in the quantum region (= > 1). in the case of large &,, the first term in

the radicand in the denominator of (1.25) may become significant. Moreover,

2hv3
c?

for Lk, « , which applies only to the short wave region (where i»,

Nis independent of &,. Finally, the requirement of maximum ~N from (1.25)
does not impose any requirements on the band width Ay and the time
constant t.

For the region where the background intensity is described by the

classical Rayleigh — Jeans formula (zr « 1) we have from (1.25)

FV Awi
Nee el

e= Van clyVko’ (1.26)

x ; hy 2hv*

and in the quantum region (+ > l and k,«— ) we have

N, = FV Awl (1.27

7 7 f
Vin c ly 24v? . )

Let us again return to Figures 18 and 19, where the dashed line marks
the limit of the two regions for &,~1. In the general case, the first and
second term in the denominator of (1.25) are equal for /,— t . If kl,

v
the boundary between the quantum and the classical region is markedly
shifted in the short- wave direction.

Quantitative estimates based on (1.25) — (1.27) lead to a definite con-
clusion regarding the optimum frequency range: the decimeter range
of wavelengths, where the radio background is minimum
(à~ 10—50 cm), ensures the maximum signal/noise ratio for
Sky surveys during a given time rv .*

In addition to being easy to detect, call signals should contain a minimum
quantity of information which will label them as artificial signals. The
fundamental differences between signals of natural and artificial origin
have not been defined yet. These differences, however, are reflected
mainly in the information content of the signals, and not in their shape.
Transmissions, and even call signals, should carry certain informa-
tion which is absent in the radiation generated by natural processes.

Another question to ask is, shall we be able to understand the com-
munications received from civilizations whose age and evolution are

* The latest observations of a new type of object — pulsars — indicate that there exists still another type of
noise in ultra-long-range transmissions. This noise, attributed to fluctuations of the refractive index of
the interstellar plasma, makes the signal disappear for long stretches of time. This effect has been poorly
studied at this stage. Unlike the background radiation, this is a multiplicative noise, and it will probably
shift the optimum frequency range toward shorter wavelengths.

EXTRATERRESTRIAL CIVILIZATIONS

substantially different from those of our civilization? There is clearly
room for understanding if a single common language can be devised.
The uniform structure of the Universe and the universality of the laws
of nature in different places and different times, as they emerge from
observational data, seem to provide this common language.

We are now in a position to summarize our conclusions regarding
call signals. Empirical considerations show that the quantity of infor-
mation needed to label a signal as artificial should contain more than
10 and less than 100 bits:

10« / « 102. (1.28)

Since the laws of nature are universal, the best policy would be
to transmit a certain combination of digits as a call signal of minimum
information content. For example, only 60 bits are required to transmit
in hexadecimal binary code the first eight primary numbers, their sum,
and the space signal between successive transmissions: 000001, 000010,
000011, 000101, 000111, 001011, 001101, 010001, 111011, 000000, ... which
stands for 1, 2, 3, 5, 7, 11, 13, 17, 59, 0,... A periodically repeated
transmission of this kind will leave no doubt whatsoever regarding its
artificial origin.

There is a great variety of different call signals. Measurements of
electromagnetic radiation record the following parameters: the two
spatial coordinates of the source, the time of observation, the frequency,
the intensity, the degree of linear polarization and its position angle,
the degree of circular polarization and its position angle. In principle,
a change in any of these parameters as a function of a change in any
other parameter may be regarded as a source of information. The dif-
ferent call signals are conveniently divided into two groups: transient
call signals and stationary (or steady-state) call signals. Transient
call signals involve a time variation in any of the above parameters
(e.g., the binary code can be transmitted by altering the sense of cir-
cular polarization). Stationary call signals involve a regular variation
of one parameter as a function of another, irrespective of the time
factor. For example, the variation of the sense of circular polariza-
tion as a function of frequency may contain the minimum quantity of
information (1.28).

It is not clear at present which of the different transmission tech-
niques is the most effective. Therefore, no exact criteria are available
for analyzing the parameters of suspicious sources.

Let us consider still another possibility of searching for call signals.
In all likelihood, only a minor fraction of the transmitter power is used
up in sending special call signals. Is it not possible to use certain
general properties of the transmitted information as a built-in call
signal? If the transmission covers a very wide frequency band, the
averaging effect may increase the measurement sensitivity several
orders of magnitude compared to the sensitivity of narrow-band mea-
surements without averaging. Thus, in radiometric measurements
of the mean source power, the signal/noise ratio increases by a factor
of n= V Avt compared to its value in measurements without averaging.
Therefore, to search for a source transmitting in a band Av, we need
an antenna with 1/n the effective area needed for receiving information

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

from the same source. As an example, if the information is transmitted
in a band of 10!! Hz and the averaging time is 10 sec, we have n = 108,

Let us now consider the problem of optimal transmission of information.
The main question is how the transmitter energy should be distributed over
the spectrum to ensure the maximum transmission rate. The fixed factors
to be considered are the noise intensity spectrum /, and the total energy
flux from the transmitter per unit surface area at the Earth, F. Problems
of this type /37/ are solved by varying relation (1.15) under the fixed con-
ditions. The optimum source spectrum is found to be

Fy= ma — H9, (1.29)

where a is determined from the condition [ F,dv=F. Smaller values of a

correspond to larger values of F. Seeing that by (1.24) Q,A, 2 a2, we find

2h t 1
F< | —m——-——]. tres)

and since F,20, we have
A
Zak > T (1.31)

where T, as before, is the effective background temperature of frequency v.
We thus reach the following conclusion: the optimum transmission
range corresponds to that part of the spectrum where
the effective background temperature is minimum.*

For the background electromagnetic radiation from outer space, this
region corresponds to the frequencies where the so-called relic radiation
prevails, i.e., the radiation described by Planck's formula with T~ 3°K.
This range covers the spectrum from submillimeter to decimeter wave-
lengths, with a background intensity maximum near 4 —1.7 mm (see
Figures 18 and 19).

A more definite shape of the source spectrum can be derived using
the dependence of A, and Q, on frequency. Let us consider two possible
cases.

1. A,2A4v3?, Q,=Q= const. As we have noted in $3, this case cor-
responds to the limitations imposed on the largest possible antennas,
provided that the relativo surface finishing accuracy is approximately
the same at all wavelengths. The shape of the spectrum F, depends on
the parameter F (Figure 21). For small F(i.e., low-power transmitters),
the maximum F, corresponds to the minimum background intensity in
the decimeter range, i.e., it coincides with the best frequencies for
callsignals. For high F, the transmitter spectrum is broader, and,
if the background radiation is negligible, we have

2hv?
A
Py = gigs + (1.32)

e ^ i

* The entire range, however, may shift toward shorter wavelengths due to the factors mentioned in the
footnote on p.47.

EXTRA TERRESTRIAL CIVILIZATIONS

This spectrum is characterized by a plateau in the low-frequency region
and an exponentially falling branch (no maximum) at high frequencies.
The shoulder is associated with information losses due to quantum fluctua-

tions of the signal.

6 7 8 9 7E I2 13 7 15 IP gunz

FIGURE 21. Energy distribution in the spectrum of an
artificial source for Ay cx v^?, Qy — const.

24-2 1
2. A,—A- const, Q, — 2 . This case corresponds to measurements

with a single antenna, which receives the entire spectrum of the signal:

2h 1 1
Fyne (= -—R— J (1.33)
e ^ -] eti

. A 0 kR{T+AT)
For small F, we may write Xa 70 oe AT «T. Then

RAT (hvy? eT
F, =27 (zr) ACE (1.34)
(, RT j)
hv 2kAT , x
For 3r € 1, the spectrum has a plateau, F,— ao then the flux increases,

reaching a maximum at LA ~1.1 (for T = 3*K, this corresponds to

A = 4,8 mm), and then falls of exponentially. The maximum F, is a factor
of 2.7 greater than the plateau value.

As F increases, the background limitations become progressively less
significant, and the spectrum width increases. At very high transmitter
powers, the distribution shows a plateau on the low-frequency side and
falls off exponentially (no maximum) at high frequencies:

2hv 1
F,-——74- i - (1.35)

The expected spectrum curve (qualitative picture) for various F is
given in Figure 22.

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

Let us now consider the general properties of information transmission
in order to find some built-in criteria for a preliminary selection of radio
sources and a search for call signals.

1. A significant part of the spectrum of an artificial source invariably
falls in the radio frequency range, with a maximum at the frequencies

corresponding to the minimum
background intensity (the short-

19^ wave part of the decimeter range)
" or in the millimeter range. A
5 Spectrum with a maximum at
2 decimeter wavelengths or with a
plateau in this range and a maxi-
I fy dT ST mum at millimeter wavelengths
2 provides a tentative criterion for
7 the selection of suspicious objects.
a 2. Minimum angular size of
-7 the suspects (radio sources) may
-2 also be regarded as a very strong
“J tentative criterion.
74 3. Measurements of other
-5 astrophysical parameters of the
-6 Source in other spectral regions
-7 can also be used for preliminary
DRE: - selection (circular polarization,
E78 IION INJ 75 n e optical and radio lines, optical
identification, X-ray emission, etc.).
FIGURE 22. Energy distribution in the spectrum of an In this respect, the search for
artificial source for Ay = const, Qy ~v~? artificial sources is virtually

coincidental with the general trend

of modern observational radio
astronomy. It is probably for this reason that the discussion of the pos-
sible tentative criteria for the identification of artificial sources left a
profound imprint on radio astronomical work. Thus, during the 1964
discussions surrounding the program of search for extraterrestrial
civilizations /38/ it was first suggested that artificial sources should
have a spectrum with a maximum at decimeter and centimeter wave-
lengths, minimum angular size, and definite variability with time.
CTA-102 was mentioned as a probable suspect meeting these criteria.
In the years that followed, the relevant properties were discovered for
a number of sources, CTA-102 included. New sources with radio emis-
sion concentrated mainly in the decimeter range were discovered. One
of the most remarkable objects in this respect is the source 1934 — 63
(coordinates a = 19434™M488.9, ò= —63°49'42" (1950)) /39/. Figure 23
shows the spectrum of this object, with a maximum around A= 21 cm.
Figure 24 is a photograph of the sky area showing this source. A galaxy
with a bright star-like nucleus is observed at the same position, and it
is joined by a hardly visible bridge to another star-like object.

A striking example of a source with a flat plateau spectrum and a
probable maximum in the millimeter or submillimeter range is 3C 273B
(see Figure 14). This and a number of other sources have extremely
small angular size (less than 0".002) and their radio flux is highly variable.

EXTRATERRESTRIAL CIVILIZATIONS

1954-39

7 8 I
lg Y Hz

FIGURE 23. The spectrum ofthe radio source 1934—63.

FIGURE 24. The sky area around the radio source 1934— 63.

As we have already noted, these observational results are inconsistent

with the synchrotron radiation mechanism generally used for radio sources.
Calculations show that this mechanism will fail to generate the observed
power in sources of such small size. Therefore, processes associated
with collective coherent emission (plasma oscillations /28/, stimulated

I. ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

emission /29/) are currently invoked to explain the observed effects.
In this connection, note that the radiation from an artificial transmitter
is a typical example of coherent emission.
The discovery of the anomalous strong line radiation at wavelengths
near 18 cm is another important factor which has bearing on our problem
/40/. The lines at 1612, 1665, 1667, and 1720 MHz are the splitting
components (A-doubling and hyperfine structure of the lowest energy
level of the hydroxyl molecule OH. Observations reveal the existence
of an unusually powerful radiation in these lines (especially at 1665 and
1667 MHz) from regions of very small angular dimensions inside ionized
gas clouds. Figure 25 is a photograph of one of these nebulae (NGC 6334); the
Squares markthe regions of anomalously strong monochromatic radiation /43/.

S 2

FIGURE 25. Nebula NGC 6334 showing regions of OH line emission.

The very existence of OH molecules in H II regions, where the temperature
is around 10,0005, is in itself a highly surprising fact. Moreover, this
emission has quite unusual properties. The angular size of the emitting
regions is less than 0",002 (linear size less than 4 a.u.). We can therefore
only give an upper bound estimate of the effective temperature at the peak

of the line profile, which turns out to be over 102 degrees. At the same time,
the unusually narrow line profile (less than 400 Hz in some cases) points

to a temperature not exceeding 10°K. This relationship between intensity
and line width is possible only in nonlinear emission processes, not

EXTRATERRESTRIAL CIVILIZATIONS

unlike the generation mechanism of molecular masers and lasers in the
laboratory. Further measurements of the interstellar hydroxyl lines
revealed an almost 100% circular polarization of the strongest compo-
nents; in some cases, strong linear polarization is also observed,
Some of the lines show pronounced variation of the component intensities
from day to day. Figure 26 is the profile of the 1665 MHz line of the
nebula W 49 in linearly polarized, right-hand polarized, and left-hand
polarized radiation /41/. Figure 27 shows the profile of the same line
of the nebula NGC 6334 on different days /42/.

130
Z7

Circular
polarization

Linear > 20
polarization Z E

-u pad - EALLA A ba.

-4 1 1.40 1 1 dt i
26 M 22 208 B wIPO0E E6420 6 i 22 0B E KEDE EEO
Radial velocity, km/sec

FIGURE 26. The 1665 MHz line profile of W 49 for linearly polarized, right-hand polarized, and
left-hand polarized radiation.

-20 6 10 i— 2 6 W -2 -5 -0 -§ 0 5 7
Radial velocity, km/sec

FIGURE 27. The 1665 MHz line profile of NGC 6334 according to observations on different days.

I, ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

Theoretical estimates of the possibilities of population inversion of
the energy levels in atoms and molecules corresponding to radio-frequency
transitions show this to be a most likely event under the natural conditions
in the interstellar medium /43/. Nevertheless, the exact mechanism of
this stimulated emission is not clear today.

It is, however, important to remember in our search for call signals
that the natural conditions in the interstellar space may greatly simplify
in this respect the problem of creation of ultra- powerful narrow-band
radio generators.

$7. THE PROGRAM OF SEARCH
FOR SUPERCIVILIZATIONS

In the preceding sections we tried to justify our thesis according to
which only the search for signals and signs of activity of supercivilizations
can be carried out with useful results in the next few years. Overa
period of some 10 years, astronomers can collect enough information
about all the brightest sources in all the regions of the electromagnetic
spectrum. This task is coincidental with the main trend of astro-
physical research today. However, now is the time to propose some sort
of a specific program for a search for artificial sources. There are
reasons to believe that transmission of information is one of the basic
conditions of existence for supercivilizations. We should therefore
develop a special program for the detection of call signals accompanying
these transmissions. Our analysis shows that the most likely frequency
range for the call signals can be identified with fair certainty. The other
parameters (frequency band, length of transmission, polarization, etc.),
however, are very difficult to guess at beforehand. If the transmitted
quantity of information is very large, we should naturally expect a very
wide band transmission, so that we will have to look for artificial sources
among a multitude of natural radio sources. It is quite probable, however,
that the low-intensity wide-band transmission carrying the bulk of infor-
mation is accompanied by powerful call signals generated at fixed fre-
quencies in a very narrow band or in the form of very brief and yet
powerful pulses.

Let us list the main directions of research which are of the greatest
interest from the point of view of the search for call signals:

1. Sky surveys at 3, 10, 30, 100, and 300 microns and especially at
1, 3, 10 mm and 3 and 10 cm aimed at discovering at least 100 of the
brightest sources in each frequency range.

2. Detailed studies of the properties of quasars and other "suspicious"
objects.

3. Search for anomalously powerful monochromatic radio sources (like
the hydroxyl line emission) in the decimeter range.

4. Search for pulse signals of interstellar origin in the same range.*

5. Search for monochromatic signals of variable frequency in the
same range.

* The search for these signals began in 1967, with the discovery of pulsars. Among the known sources of
this type, however, there are still no indications of artificiality.

EXTRATERRESTRIAL CIVILIZATIONS

The currently available information on each of the five items above
is pitiful and negligible compared to what could have been collected with
a properly planned utilization of modern means. The preliminary
criteria discussed in the previous section may prove to be of great help
in the preliminary sifting through of "suspects." The angular size
criterion is particularly useful. The angular dimensions can be accurately
measured with a radio interferometer. Radio interferometers with a
base of the order of the Earth's diameter are currently available for
the centimeter and decimeter wavelengths /44/. In the near future,
one of the antennas will probably be mounted on an interplanetary space-
craft, thus giving a radio interferometer with a base comparable with
the dimensions of the Earth's orbit. Other promising directions include
the estimates of the maximum linear dimensions from the time variation
of one of the source parameters (e.g., radiation flux or polarization).
Since the velocity of light is finite, the radiation of the entire object

can be observed to change simultaneously in a time t only if t», where

r is the radius of the object. Suppose that the quasar 3C 273B is an object
of mass M ~ 10? solar masses and its radius is greater than the critical

(gravitational) radius r= oe = 3-108 cm (G is the gravitational constant).

We thus come to the conclusion that the brightness of this source cannot
change faster than with a period of T~ z = 10? sec. Any discovery of

faster light variation would point to a smaller mass and radius of this
object.

Preliminary selection using the tentative criteria is a necessary,
though not sufficient, stage of the general search procedure. Once a
sufficient number of "suspects" have been selected, we have to start
looking for "meaningful contents" in the radiation from these objects.
This work, supported by parallel theoretical analysis of the various
alternatives, will help to improve the future search program. In par-
ticular, we hope that significant information on the parameters of
quasars and their time variation in various spectral regions will be
accumulated in the course of the international program launched in
1966 /45/.

At present, we have no theory to enable us to assess the presence
or the absence of meaningful information in the received signals. Man
is the only suitable candidate for making decisions in this direction, and
we are thus inevitably faced with the difficulties of subjective approach
to the search program. This approach, however, will not be entirely
arbitrary. A certain measure of objectivity willbe derived from the observed
universality of the laws of nature and their constancy in space and time.
The universal laws of nature can be used as a common basis of under-
standing with other civilizations and, in particular, enable us to develop
an objective search program. In principle, we can probably devise a
procedure and build an analyzing machine for the comparison of the known
universal laws of nature (mathematical relations in the simplest case)
with any information received from outer space. In our opinion, this
problem is definitely solvable, at least as far as the search for call
signals is concerned.

I, ASTROPHYSICAL ASPECT OF SEARCH FOR SIGNALS

Bibliography
1. Shklovskii,I.S. Vselennaya, zhizn', razum (Life and Intelligence
in the Universe) 2nd Ed.—''Nauka." 1965.
2. Cameron, A. (Editor). Interstellar Communication.— New York.
Benjamin. 1963.
3. Shklovskii,I.S. and C.Sagan. Intelligent Life in the Universe.—
Holden Day. 1966.
4. Vologdin, A.G. Pervye shagi evolyutsii (The First Steps of
Evolution). — Literaturnaya gazeta, 1 February, 1967.
5. Baranov, V.I.— Astron. Zhurnal, Vol. 43:1074. 1966; Fisher, D. E. —
20th IUPAC Congress, Moscow. 1965.
6. Gerling, E.K., V. A. Maslennikov and Ii. M.Morozova.- Ibid.
7. Nablyudatel'nye osnovy kosmologii (Observational Principles of
Cosmology). Collection of articles. — Mir. 1965.
8. Burbidge,E.M. Quasi-stellar Objects. — San Diego, Univ. of California,
California. 1967.
9. Lyapunov,A.A. — Problemy Kibernetiki, No. 10: 179. 1963.
10. Kardashev,N.S. and G. B. Sholomitskii.— Astr. Tsirk., No.336.
1965.
11. Rose,D. and M.Clark. Plasmas and Controlled Fusion. — Cambridge,
Mass. MIT Press. 1961.
12. Andrillat,Y. and M. Andrillat.— Publ. de l'Observatoire de Haut
Provence, Vol.7, No.11. 1964.
13. Dibai,E.A. and V.I. Pronik.— Astr. Tsirk., No.286. 1964.
14. Ozernoi,L.M. and V.E.Chertoprud.-— Astron, Zhurnal, Vol. 43:20.
1966.
15. Geyer,E.—Zs. für Astroph., Vo1.60:112—114. 1964.
16. Kinman,T., E.Lamla, and C. Wirtanen. — Contr. from Lick
Observatory, No.225. 1966.
17. Sandage, A., J.Westphal, and P.Srittmatter.— Ap. J.,
Vol.146:332, 1966.
18. Hoerner,S.von, — Ap.J., Vol.144:483, 1966.
19. Adgie R., H.Gent, O. Slee, A. Frost, H.Palmer,and
B.Rowson.- Nature, Vol.208:275. 1965.
20. Cohen, M., E.Gunderman, M.Hardebeck,and L. Sharol.—
Sky and Telescope, Vo1.34: 143. 1967.
21. Kellerman,K. and I. Pauling— Toth.— Ap. J., Vol. 146. 1966.
22. Berge,G. and G. Seielstad.-— Observations of the Owens Valley
Radio Observatory, No.9. 1966.
23. Low, F.— Ap. J., Vol.142:1287. 1965; Ap. J., Vol. 150. 1967.
24. Friedman,H. and E.Byram.~7 Cospar Symposium, London,
24—28 July. 1967.
25. Shklovskii,I.S.— Astron. Zhurnal, Vol. 42:893. 1965,
26. Sandage, A. and W. Muller.— Ap. J., Vol.144:1240. 1966.
27. Veron, P.— Ann. d'Astrophys., Vol. 29:231. 1966.
28. Ginzburg,V.L. and M. M.Ozernoi.— Ap. J., Vol.144:599. 1966.
29. Zheleznyakov,V.V.— ZhETF, Vol.57:570. 1966; Astron. Zhurnal,
Vol.44, 1967; Kaplan,S.A.— Astrofizika, Vol.2:409. 1966.
30. Byram,E., T.Chabb,and H. Friedman.-— Science, Vol1.152: 66.1966.
*31. Moffet, A. — Annual Review of Astronomy and Astrophysics, Vol. 4.1966.
32. Williams, P.— Observatory, Vol. 86:67. 1966.

33.

34.
35.
36.

37.

38.

39.
40.

41.
42.

43.
44.

45.
46.
47.

EXTRATERRESTRIAL CIVILIZATIONS

Dyson, F.— Science, Vol.131:1667. 1960; Perspectives in Modern
Physics, "Thoughts on the Search for Extra Terrestrial Technology."
N.Y. Interscience Publishers. 1966.

Sagan, C. and R. Walker.— Ap. J., Vol.144:1216. 1966.

Zel'dovich, Ya. B. — UFN. Vol. 89:647. 1966.

Rocchia, R., D. Rothenflug, D.Boclet, G.Duero s, and
Y.Labeyrie.—7th Int. Symp. Space Explor., Vienna, 11—17 May.
1966.*

Lebedev,D.S. and L. B. Levitin. Perenos informatsii elektro-
magnitnym polem (Information Transmission by Electromagnetic
Fields). — In Sbornik: "Teoriya peredachi informatsii, Problemy
peredachi informatsii," No. 16. 1964.

Vnezemnye Tsivilizatsii (Extraterrestrial Civilizations). Proceedings
of a Conference, Byurakan, 20—23 May 1964.—Izd. AN Arm SSR.
1965.*

Kellerman,K.— Austr. J. Phys., Vol.19:195. 1966.

Weaver,H., D.Williams, N.H.Dieter,and W.Lum.- Nature,
Vo1.208:29. 1965.

Palmer,P. and B.Zuckerman.—HRAP, Vol.124. 1966 (preprint).

Dieter,N.H., H.Weaver and D. William s.— Sky and Telescope,
Vol, 31:132. 1966.

Varshalovich, D.A.— ZhETF Letters, 4(5):180. 1966.

Kaidanovskii,N.L. and N.A.Smirnova.-— Radiotekhnika i
Elektronika, Vol.10:1574. 1965; Sky and Telescope, Vol.34:143.
1967.

Symposium IAU, No. 29. Byurakan, May 1966.

Sholomitskii,G.B.- Astron. Zhurnal, Vol.44:939. 1967.

Reddish, V. C.— Vistas in Astronomy, Vol. 7:173. 1966.

* [See footnote on p.11.]

5780

Chapter II

THE EFFECT OF THE SPACE MEDIUM ON
THE PROPAGATION OF RADIO SIGNALS

The search for signals of extraterrestrial civilizations is closely
associated with a painstaking analysis of the radio waves received from
sources in outer space. The propagation of radio signals in the outer
space is therefore one of the main topics in our analysis.

The outer space (including the interplanetary, the interstellar, and
the intergalactic medium) is characterized by extremely low density of
matter. The effect of the space medium on signal propagation is there-
fore also very low. However, because of the tremendous distances that
the signals traverse before reaching the observer, the weak effects can
build up to alarming magnitudes. The integrated cumulative effect may
introduce significant distortions into the signal characteristics.

A detailed analysis of the propagation conditions clearly requires
knowledge of the basic parameters of the space medium: density of
matter, inhomogeneity of the medium, temperature, magnetic fields.
These data (especially for the intergalactic medium) are fairly uncer-
tain at this stage. Moreover, no detailed analysis can be carried out
without giving consideration to the particular characteristics of certain
limited regions of space through which the radio waves travel. For
example, the conditions of propagation of radio waves in the Galactic
plane are substantially different from the conditions of their propaga-
tion toward the Galactic pole. Individual objects (e.g., a dense cloud
of ionized hydrogen) intercepting the line of sight may introduce sig-
nificant distortions into the signal compared to the "average" propaga-
tion conditions.

We are unfortunately in a position to give only some general limiting
estimates of the effect of the space medium on radio propagation.

Fairly numerous studies are available, dealing with such estimates,
We will therefore consider the main conclusions pertaining to the "inter-
ference" from the space medium.

Absorption is one of the leading factors which affect the propagation
of radio waves in a material medium. From the classical point of view,
the absorption of radio waves can be described as oscillatory pumping
of electrons by radio waves, which subsequently lose the extra energy
through collisions with protons. The absorption coefficient per unit
path length is expressed by the relation

p= I=? coll (2.1)

cn eff’

EXTRA TERRESTRIAL CIVILIZA TIONS

where n is the refractive index of the medium, ve is the effective number
of electron-proton collisions. The expression for the refractive index can

be written in the form
4ne?N.
ay 1- S. (2.2)

Here N, is the electron concentration of the medium, w is the frequency

of the radio waves.
For the propagation of centimeter and decimeter waves in the rarefied

interstellar and intergalactic medium, when dee « 1, the absorption

coefficient is expressed in the form

0.58N? 4.3 -10°T
p=—,— In (x. (2.3)

T^v? oh

where T is the temperature of the medium.

The optical thickness for absorption (which measures the amount of
absorption) t=pl (l is the path length of the radio waves in the medium)
is proportional to the measure of emission N2-1, where N?is the Square
of the mean electron concentration in the medium along the entire path.
The measure of emission for distances comparable with the size of the
Galaxy lies between the limits 6-10 < N2 «6.1070,

The optical thicknesses t for various measures of emission and
frequencies o are listed in Table 2.1.

TABLE 2.1. The optical thicknesses t (in the Galaxy)

cg
Pr 6-102» | 6-10" | 6-108
1010 4.1075 4-107* 4.107
10° 4:107? 4.107% 4.1075
10* 4-107 4-107 4-10°
10? 40 4 0.4

The table shows that radio waves with frequencies w > 10? Hz propagate
virtually without absorption in any direction in the Galaxy. (The magnitude
of absorption is determined by the factor e-t.) Transmission at lower
frequencies is obstructed by strong absorption, especially in the direction
of the Galactic plane, where Ni.[ reaches its maximum values.

The refractive index (2.2) also determines the dispersion effects which
distort the transmission. We should distinguish between two effects:
the phase shift of the spectral components of the signal due to dispersion
in the medium and the "lag" of the quasimonochromatic group components
which transmit the signal energy.

II. EFFECT OF SPACE MEDIUM ON PROPAGATION

To illustrate the difference between these two effects, let us consider
the transmission through space of a train of pulses, pulses of length P
following one another at intervals of the same length (Figure 28).

A pulse of length P (and any train of such pulses) can be expanded
into a spectrum (a Fourier integral). The train of pulses shown in
Figure 28 has a spectrum which covers a frequency interval of width

2x
AQ >.

FIGURE 28. Undistorted pulses.

In dispersive media (n#1), the phases of the individual spectral

components of the signal propagate with different velocities up, =F"
Therefore, the phases of different components acquire a relative shift
and the resultant combination at the receiver gives a distorted pulse
shape /1/. Depending on the characteristics of the propagating medium,
the pulse either "contracts" or "spreads."

We have considered the propagation of high-frequency pulses. Each
pulse was "filled" with monochromatic radiation of a high frequency o.
This treatment is valid only for a steady-state transmission of a high-
frequency signal (when wP>>1). This means that the pulse length
accommodates a considerable number of periods of oscillations of frequency
w. Inthe opposite case (wP < 1) the pulse involves a macroscopic variation
of the intensity of a low-frequency field in the medium.

Suppose that the pulse train (see Figure 28) is generated in the following
way: a certain radiation source with a sufficiently wide continuous spectrum
is periodically obscured by a screen. The pulse train emitted into space
will then have a wide-band "filling," possibly not unlike noise (thermal
"noise').

Let the frequency band of this radiation be Aw, and the spectral density
E(w). From the energy; point of view, each pulse is a collection of quasi-
monochromatic groups E(e)óe, where ôw is a very narrow "quasimono-
chromatic" band in the spectrum. The integrated effect of all these group
intensitíes gives the height and the length of the pulse. Quasimonochro-
matic wave groups propagate with a group velocity v,,=cn(w). For ionized
space media, this velocity decreases with decreasing frequency. Asa
result, over sufficiently long distances, the high-frequency wave groups
precede the low-frequency groups, and a characteristic "time sweep"
of the spectrum is obtained. The importance of this effect in connection
with solar radio bursts with frequency drift was discussed in /6/. A
similar effect relating to the propagation of radio waves in interstellar
media was discussed in /2/. The problem was also considered in /3,4/. *

* This effect was first discovered in observations of pulsars — pulsating radio sources.

EXTRA TERRESTRIAL CIVILIZATIONS

The delay of a wave group of frequency o, relative to a wave group of
frequency wœ: can be found from the relation

L l
Atmo) T zy a via)

If the frequencies oi, o» are far from the critical frequency o,, at
which n(o,,)=0 (this condition can be written in the form n« 1), the delay

is expressed by the formula

Qne? o?— 2 —
At (à — 95) ~ — —- Me. (2.5)
195

This expression is conveniently rewritten taking w, and we in the form

Aw
aSo t- >

Aw
0: = 0- >

where w is the mean frequency of the signal. Then

At (Ao) = ZE AP w, (2.6)

The delay At for space media for various Aw and wo is listed in
Table 2.2. We see from the table that the delay may reach considerable
values. What does this lead to ?

‘Vo answer this question, we have to consider the conditions of reception
of the signal shown in Figure 28. Suppose the receiver band Awrec > Aw,
i.e., the receiver is capable of receiving the full intensity of the entire
spectrum of the signal Ae. For simplicity, we take E(w)= const in the
entire frequency band Aw. Clearly, if A/(Ao) «P, no significant distortions
will be introduced in the received signal. If, however, At~P (Figure 29,a)
the signal is markedly distorted. The high-frequency spectral compo-
nents of the signal are the first to be received. The low-frequency
components are delayed and the signal "spreads."

If AC» P (Figure 29, b), the pulses are completely blurred into a
continuous emission from the source (its intensity is much less than the
peak pulse power).

The true signal shape in principle can be restored by an appropriate
correction in the receiver or in the processing stage. The unfortunate
fact, however, is that we do not have the actual numerical values of the
parameters of the media propagating the pulses from outer space.

The periodicity can be "caught" (for A» P) by narrowing the receiver
band Ao,« to such an extent that Af (Awe) <P. This procedure, however,
will lead to substantial losses of the receiver sensitivity (which is

2
proportional to (==) , a factor not to be trifled with in the reception
of signals from outer space. Moreover, the narrower receiver band
will have an adverse effect on the rate of information transmission (see

Chapter III).

Il. EFFECT OF SPACE MEDIUM ON PROPAGATION

TABLE 2.2. Lag time for wave groups, in sec (Galaxy, Metagalaxy)

5-10? (the limit for interstellar distances)

Aw = 0 5o, | AQ = 0.104 | Aom 1070 | 5-7,

1010 0,25 5-107? 5-104 5. 107*

09 c 25 5 5.10? 5.107*
108 2,5. 10? 500 5 5-107
107 2,5 - 105 5. 10* 5.107 5

5.10!

ào-05, | Awa 0 19, | AQ 10730, | soe 10-50,

1010 2.5 05 5.107? 5.10?
10° 250 50 0.5 5.107
108 2.5. 10* 5-10? 50 0.5
10? 2.5. 10° 5-105 5.10 50

2:10” (the limit for intergalactic distances)

Aw = 0.50 | sam 10a, | Anma | Ao~ 10770,
1010 10? 0.2 2.107? 2.107?
109 10 20 0.2 2.10?
10 105 2-104 20 0.2
107 10? 2-108 2-104 20
7
2
rF r= reg
It 1 q n eue: Oaa
Z
CEN Bud E
a
i
ra "o3 r3 r3
AES eee
poe nav Pd
poe Do Soe a E EL
b

FIGURE 29. Distorted pulses:

a) the case Af ~P; b) the case V» P.

TABLE 2.3. Minimum pulse length (seconds)

Nyt
OE 5-10* 5-10 5-107 2-107
10" 25-10? 8.10? | 25.107 5-107
10? 8-10* | 25.10? 8.1077 15-107?
10* 25-107? 8-107 0.25 0.5
10! 8-107 25 8 15

ee eee - V. E —— ——

EXTRATERRESTRIAL CIVILIZATIONS

The group lag effect imposes certain restrictions
on the permissible pulse length P. To avoid the
highly undesirable significant distortions which
we described above, we have to ensure the inequality
Al(Ao) «P. The minimum values of P prescribed by
this requirement are listed in Table 2.3.

Radio waves propagating over large distances
in the intergalactic medium may also show a "red
shift." The red shift has an "unfavorable"! effect,
lowering the frequency of the propagating radiation.
The distortion effects are therefore enhanced for
propagation over very large distances.

The distortions introduced by the space medium
into other types of radio signals (frequency or phase
modulated signals) should be considered separately.

Analysis of the data in Tables 2.2 — 2.3 stresses the advisability of using
the shortest wavelengths in the radio spectrum for long-range interstellar
communication.

The effect of radio wave propagation conditions in the space medium
and in the Earth's atmosphere on the apparent angular size of the radio
source has been discussed in /5/. If the propagating medium is in-
homogeneous, the wave front is distorted on passing through this medium
(Figure 30). The amount of wave front distortion is determined by the
deviation of the wave phase from the unperturbed value. Proceeding from
some model considerations (e.g., the size of inhomogeneities, the
mechanism of wave scattering by the inhomogeneities, etc.), we can
arrive at an average statistical estimate of the integrated distortion
acquired by a wave on passing through an inhomogeneous layer of a given
thickness. The mean square phase deviation from the unperturbed value,
$?, was calculated in /5/ using the expression

FIGURE 30. Wave front dis-
tortion.

"hi. d —À,
ga tan, (2.7)

where / is the path length of the wave in the inhomogeneous medium, d is
the mean inhomogeneity size (it is assumed that d>), a is the wavelength,
An? is the mean square fluctuation in the refractive index of the medium.
Using the geometrical optics approximation, we can obtain an expression
for the deviation angle o of the beam from the original source — observer
direction, We have

8? = Anh Bn?, (2.8)

The spreading of the angular diameter of the source to c will be observed
in the far zone of the scattering region (i.e., at distances of the order

Re i>). Tables 2.4 and 2.5 show that for $?7 1, this condition is not

Satisfied on the Earth. The Earth-bound observer is located in the near
Scattering zone of the space inhomogeneities. In this case, the distortions
introduced by the medium in the size of the source are determined by the

Il, EFFECT OF SPACE MEDIUM ON PROPAGATION

relations between V 5^, the angular size of the inhomogeneities
ya? (a - f) and the antenna size D.
The values of Yo’, Và? for various media are listed in Table 2.5.

V8? > Vi? Dod

FIGURE 31. Deviation of light rays propagating through an
inhomogeneous medium.

If V&^» Và, an antenna for any diameter will receive rays which have
covered a distance greater than the correlation radius (as determined by
the mean inhomogeneity size), and the source will expand to the full angular
size c (Figure 31, a).

TABLE 2.4. The parameters of space media used for the calculations in Table 2.5

Medium ] t, cm d, cm Bn? | N,

Troposphere 1.5: 10$ 0.5. 107$
lonosphere 4-107 4.5: 107 3?
Ecliptic plane 10% 4.5: 107 73? 10?
Interplanetary 4| Toward the 0.5: 1053 0.5: 1074? 20
pole
Galactic plane 6.107 4.5: 107 ^4? 3-107?
Interstellar To the Galactic 6-10” 4.510 "a? 3-107?
pole
Intergalactic 1078 4.5. 1074?

If Vo? « Và, the effect will vary depending on the relation between
D and d.

In a filled-aperture antenna of size D>d (Figure 31, b) the phase
fluctuations produced by inhomogeneities cause a loss in the effective
area and broaden the beam angle.

For antennas of size D«d(Figure 30), refraction effects are observed,
which shift the apparent position of the source throughthe refractionangle.
As a result, the source coordinates are measured with a certain error.

EXTRATERRESTRIAL CIVILIZATIONS

TABLE 2.5. Distortion of point source image

Range of wavelengths

Space medium ye where 921
yeya

V3? «ya

Interplanetary
medium

Ecliptic plane | 1.4- 1074? 4:1079A? 1075 entire spectrum] à « 16cm

0.28.42 1.7:107 93? 2.1074 à »2cm 4«8.5-10* cm

Galactic plane| 2.5.1015? | 1.7: 107 !!2? 5.1075 | entire spectrum [A< 8.5- 10*cm

Interstellar
medium

2.5.1044? | 1.2: 107A? 5:10 ° | entire spectrum |entire spectrum

1.4. 1055 A? 1078 entire spectrum |entire spectrum

Intergalactic medium 1.2:107 53?

In these calculations, the possible motion of the inhomogeneities should
be taken into consideration. If the inhomogeneity clouds move with a
certain velocity v, the source will "shimmer" with a period

d
tE (2.9)

Scattering effects thus limit the resolving power of antennas. There
are two alternatives: either the source expands V8?» V& ) or the
Scattering has an adverse effect on antenna directivity (V à? « Và?, D» d).

Table 2.5 lists the wavelength region for the various Space media
where the conditions g?>>1, Vó?« Và? are satisfied. The interplanetary
medium evidently introduces considerable distortion in the angular size
of the source.

Imm /cm Wom 7m fm A

FIGURE 32. Limiting antenna resolution.

Figure 32 plots curves of the limiting antenna resolution derived with
allowance for the effect of scattering in space media. Filled-aperture
antennas are very limited in terms of resolution. The maximum resolving
power is attainable only using radio interferometers (at wavelengths

Il, EFFECT OF SPACE MEDIUM ON PROPAGATION

shorter than 10 cm). This again stresses the advisability of using the
shortest wavelengths of the radio spectrum in observations.

In conclusion note that the "reversal" of the problem of distortions
introduced by the space medium may prove quite fruitful for astrophysical
purposes. If the true dimensions of the source or the parameters of the
variable radio signal from the source can be estimated from independent
considerations, the analysis of distortions introduced by the space medium
may provide highly valuable information on the properties of the medium
itself (material density, size of inhomogeneities, etc. ).

Bibliography

1l. Ginzburg,V.L. Rasprostranenie elektromagnitnykh voln v plazme
(Propagation of Electromagnetic Waves in Plasma). — Fizmatgiz.
1960.

2. Gudzenko,L.I. and B. N. Panovkin.—In: "Vnezemnye tsivilizatsii,'
Proceedings of a Conference. Byurakan, 20—23 May 1964, p. 68.
Izd. AN Arm.SSR, 1965.*

3. Haddock, F.I. and D. W.Sciana. — Phys. Sci. Let., 14(25):1007. 1965.

4. Panovkin, B. N. — Fifth Soviet Conf. on Radio Astronomy, Khar 'kov.
1965.

5. Kaidanovskii,N.L. and N. A, Smirnova.-— Radiotekhnika i
Elektronika, Vol.10:1574. 1965.

6. Wild,J.P., K.V.Sheridan,and A. A. Neylan.-— Austr. J. Phys.,
Vo1.12:369. 1959,

* [See footnote on p. 11.1]

Chaptev III

THE POSSIBILITY OF RADIO COMMUNICATION
WITH EXTRATERRESTRIAL CIVILIZATIONS

The topic of communication with extraterrestrial civilizations (EC)
has repeatedly cropped up in the scientific literature /2,3/ after the
pioneering work of Cocconi and Morrison /1/, who were the first to
establish the feasibility of communication with EC in the electromagnetic
Spectrum. There is no doubt that the organization of communication with
ECisanunprecedented technical problem, whose specific requirements cannot
be fully appraised at this stage. On the other hand, it seems that any
communication system, including the system of communication with EC,
would satisfy certain general requirements which follow from the general
laws of information transmission. The study of these laws is the subject
of the information theory or the general theory of communication. We will
therefore start our review with a discussion of the principal elements
of the general theory of communication, which will prove useful in the

following.

$1. ELEMENTS OF THE GENERAL THEORY OF
COMMUNICATION

Structure and fundamental characteristics of a com-
munication system

The aim of any communication system is the transmission of certain
messages. The messages may constitute text written using the letters
of a certain alphabet (as in telegraph messages) or sounded verbally
(telephone, radio). The message may also constitute an image of a
certain object (phototelegraph, television) or an algorithm to be trans-
mitted to an automatic control system. Any of these messages can be
represented as a succession of digits or as some continuous time function
x(t).
Messages are transferred by a communication system using certain
agreed signals. In the present chapter we will only consider systems
employing electrical signals.* An electrical signal is a time-variable

* In a more general treatment of communication, when we are dealing with such systems as biological
population, biological evolution, etc., the basic concepts of message, signal, and information require
a new, more precise definition. However, after an appropriate generalization of these concepts,
the basic propositions of the theory of electrical communication prove to be valid for a larger class
of communication systems.

Ill, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

electrical magnitude (voltage, current, field strength) and, like the message
itself, it can be expressed as a certain function of time. The signal re-
flects the message in the form of an electrical disturbance. The trans-
mitted message must be reconstituted from the received signal.

Source of
noise
FIGURE 33. A generalized communication system:

1— message, 2 — signal, 3 — noíse, 4 — signal + noise 5 — received message.
The part of the block diagram enclosed in the dashed rectangle is the communi-
cation channel.

A block diagram of a generalized communication system is shown in
Figure 33. The message from the information source, or the sender, is
delivered to the transmitter which transforms it into a signal sent through
the communication line. The communication line is an electromagnetic
wave channel, or, in other words, the medium propagating the signal
from the transmitting end of the system to the receiving end. The line
may comprise two conducting wires, a coaxial cable, a waveguide, or
the unrestricted part of space in which radio waves propagate. Thus,
for short-wave radio communication, this is the spherical layer between
the Earth's surface and the ionosphere. In directional radio transmission,
the communication line is the part of space inside the solid angle subtended
by the receiving antenna.

A signal propagating along the communication line may experience
distortion and may be intermixed with noise. Distortion is generally
described as those changes in the signal which are caused by known
characteristics of the system. In principle, these distortions can be
corrected, and we will not have to analyze their effects.* Noise, on the
other hand, is random and cannot be fully corrected. Random noise is of
the greatest importance for the actual performance of a communication
line.

At the receiving end ofthe line, the electrical message is picked up by
a receiver, which constitutes the original message by an appropriate
transformation of the received signal. Mathematically, the action of the
receiver is the inversion of the transmitter action. The part of the system
including the transmitter, the line, and the receiver is generally designated
as the communication channel.

Distortions experienced by a signal propagating in the interstellar medium were considered in Chapter II.
Note that when propagating in a medium with randomly changing properties, the signal experiences
random distortions which cannot be corrected. An example of such random distortions is the scintillation
of stars and radio sources. The same effect causes a definite broadening of the angular dimensions of the
sources (see Chapter ID.

EXTRA TERRESTRIAI, CIVILIZATIONS

In an ideal communication system, free from noise, the received
message is identical to the transmitted message. In real noisy systems,
however, this is never so. The degree of identity of the received and
transmitted signals characterizes the reliability of the communication
system. The reliability depends on the ratio of the signal power to the
noise power in the communication signal. Asa rule, the reliability
falls off with distance. The maximum distance over which a certain
reliability is still attainable is known as the communication range.

This parameter is naturally of the greatest importance in systems of
communication with EC.

Another important characteristic is the transmission rate of the
communication channel, i.e., the quantity of information that can be
transmitted by the given communication system in unit time. The
transmission rate characterizes the information content of the transmitted
message. However, the system does not "distinguish" between important
and trivial messages. Thus, to send a telegram consisting of 100 symbols,
the system should always meet certain fixed requirements (transmission
time, frequency band, signal power, etc.), regardless of the importance
and the content of the message. The concept of information in the
general theory of communication is therefore devoid of any qualitative
meaning, and should be treated as a pure quantitative concept.

Quantitative definition of information

How are we to define information? Consider the transmission of a
sequence of four-digit decimal numbers. These are either numerical
values of some physical magnitude or four-letter words written using
a ten-letter alphabet. Suppose we are transmitting a certain word M.
What is the information content of our message? The total number of
four-digit numbers or possible messages is N— 104, By transmitting
our message, i.e., a particular number M, we have made a definite
choice out of the available total of N=104, The number N of the
available choices characterizes the uncertainty of the outcome prior to
the transmission. This number ~N is also used to characterize the
information content of the particular message. The higher the initial
uncertainty which prevailed before the transmission, the higher is the
quantity of information contained in the message, and conversely: the
lower the initial uncertainty, the lower is the quantity of information
in the transmitted message. If the quantity of information is designated

Q, we may write

Q — Q(N), (3.1)

where Q is a single-valued monotonically increasing function of N. From
this definition we see that if V,—- N,, then

Qi = Q(N) = Q(N) = Qs. (3.2)

Ill, RADIO COMMUNICATION WITH EXTRA L ERRESTRIAL CIVILIZATIONS

The four-digit decimal number M can be expressed in a binary, ternary,
or any other number system with some base a. Then N, =a?, N,=ay,
where m is the exponent of the number M (for a whole m, this is simply
the number of digits needed to express the number in the given system).
Condition (3.1) thus takes the form

Qi=Q2 if af'=ag". (3.3)

In this form, it simply means that the quantity of information contained
in the number M is independent of the particular system used to express
this number.

'The next condition to be met by our definition of information is that if
we take different numbers expressed in the same number system, the
information content of each number will be proportional to the number of
digits (i.e., a six-digit decimal number 145876 contains double the in-
formation of the three-digit number 963 and three times as much informa-
tion as the two-digit number 25). Thus,

Q=ym, (3.4)

where y is a proportionality coefficient.

In application to the problem of information transmission through a
channel, this means that the quantity of transmitted information increases
linearly with transmission time (indeed, to transmit six digits, we need
double the time to transmit three digits). Thus, a two-minute trans-
mission is in general (other conditions being equal) more informative than
a one-minute transmission.

It can be shown that conditions (3.1) — (3.4)* define a unique function

Q —log,N =m log» a. (3.5)

This definition was first advanced by Hartley /4/ in 1928 and ithas been used
since with excellent results in the theory of communication.

The base 6 of the logarithm in (3.5) is arbitrary. The choice of this
base corresponds to the unit of information measurement. Taking b- 2,
we obtain the quantity of information Q in binary units, or bits. This
unit of information, corresponding to the lowest possible base of a number
system, may be adopted as the basic unit of information measurement.

It is widely used in applications.

Let us now determine the information content of our four-digit decimal
number M. Taking m —4 anda- 10, we find Q = 41og210 ~ 13.3bits.
Similarly, the quantity of information in a five-letter word from a 30-
letter alphabet is 5 logs30= 24.6 bits, and a text of 100 words with an
average word length of 5 letters contains about 2460 bits of information.
For a=b= 2, Q=m bits, i.e., the quantity of information, expressed
in bits, contained in a number M is equal to the number of binary digits
required to express this number in a system with a base 2 (for whole m,
naturally).**

* Since conditions (3.1) — (3.3) are not independent, any two of the conditions are sufficient for a single-
valued definition of Q, e.g., (3.1) and (3.4), (3.2) and (3.4), or (3.3) and (3.4).

If m is not a whole number, M isexpressedusing m; binary digits, where m, is the nearest whole number
to m, m>m=Q.

EXTRA TERRESTRIAL CIVILIZA TIONS

The above definition applied to discrete messages. However, a con-
tinuous function of time can be represented with any desired accuracy by
a set of discrete quantities, and this definition is therefore quite general
for the purposes of communication theory.

Let us now consider the various techniques whereby a message is
transformed into a signal.

Transformation of a message into a signal. Forms of
modulation

A signal is transmitted as a direct current, electromagnetic oscillations
of high frequency, or a periodic train of pulses. When a signal is trans-
mitted down a communication line, one of the line parameters varies in
accordance with the transmission function x(t).

Direct current is characterized by two parameters: the magnitude
and the direction of the current. By changing one of these magnitudes
in accordance with x(t), we obtain an electric signal which may propagate
along the communication line (e. g., as in the transmission of Morse-
coded telegrams). However, since direct current will propagate only
through wires, this method of transmission is of no consequence for our
problem.

The signals in radio communication are high-frequency electromagnetic
oscillations which may propagate freely through the vacuum. Sinusoidal
oscillations are characterized by three parameters: the amplitude, the
frequency, and the initial phase. By altering one of these parameters in
accordance with the message function, often called the modulating function,
we obtain a modulated electromagnetic signal of high carrier frequency.
We thus distinguish between three different forms of modulation,
corresponding to the three parameters of the carrier: amplitude
modulation AM, frequency modulation FM, and phase modulation PM
(Figure 34a). The modulated signals are demodulated in the receiver
to reconstitute the modulating function x(t), which is the message.

If the signal is transmitted by a periodic train of pulses, we obtain
four types of pulse modulation corresponding to variation of the pulse

height 4, pulse duration t, and pulse recurrence frequency w= (T is

the time between two successive pulses): these are the pulse-amplitude
modulation, PAM, the pulse-duration modulation PDM, the pulse fre-
quency modulation PFM, and the pulse position modulation PPM (Figure
34b). In a number of cases repeated modulation is used: the pulse train
is modulated by the message function, and the modulated pulses are then
used to modulate a high-frequency carrier (Figure 34c). This provides
a new modulation technique, high-frequency pulse modulation HFPM, in
which the height, length, frequency, and phase of pulses remain constant,
and only the duty cycle is altered (Figure 34d).

An important variety of pulse modulation is the transmission of coded
messages. We will consider this type of modulation after becoming better
acquainted with some properties of signals.

IIl, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

a) Carrier modulation b) Pulse moduiation

d) High-frequency pulse modulation HFPM

3 a} Z

LM. t H
FIGURE 34. Formation of electrical signals by modulation:

x (t) is the message function or the modulating function, AM amplitude
modulation, FM frequency modulation, PM phase modulation; PAM
pulse-amplitude modulation, PDM pulse-duration modulation, PFM
pulse-frequency modulation, PPM pulse position modulation.

Physical characteristics of signals

A signal canbe characterized by the following three parameters: signal
duration, the dynamic range, and band width.

Signal duration is the simplest characteristic. Its practical importance
is self-evident: the longer the signal, the longer it takes to transmit it and
the longer the lines remains engaged.

The dynamic range is defined as the ratio of the maximum instantaneous
signal power (the so-called peak power) to the minimum signal power.

The dynamic range is measured on a logarithmic scale and is expressed in
decibel. One decibel (1 dB) is equal to 0.1 on the logarithmic scale;
therefore if n is the ratio of the measured quantities on the linear scale,
the same ratio in dB is equal to 10 logn. Signals where the peak power

is double the minimum power have a dynamic range of 3 dB; a dynamic
range of 10 dB corresponds to a maximum-to-minimum ratio of 10, 20 dB
to a ratio of 100, 30 dB to a ratio of 1000, etc.

The choice of the minimum signal power is determined by the noise
level. To ensure a reliable reception, the minimum signal power should
exceed by a certain factor the mean noise power P, (Pmn=aP,). High-
quality transmission of speech by amplitude modulation requires Pmm
exceeding the mean noise power by 60—70 dB. The quantity Pam=aPn

EXTRATERRESTRIAL CIVILIZATIONS

isknownasthe threshold signal power. Thedynamicrange, relatedto
the threshold power, is often replaced by the ratio of the mean signal power
to the mean noise power P,/P,. This ratio is briefly called signal-to-noise
ratio or signal/noise ratio. Both the dynamic range and the signal-to-noise
ratio characterize the signal power relative to the noise power, and not
the absolute power. What are the factors determining the threshold power?
Suppose we wish to transmit a certain message, which expresses the
value of the function x(/) at the time 4. We may use one of the pulse
modulation systems, e.g., the pulse-amplitude modulation, and send a
pulse of height x(fo) =x along the communication line. In the case of an
ideal noise-free channel, this pulse is received without distortion at the
receiving end of the communication line, and the original message x(t)
will be recovered from the pulse amplitude x). In a real channel, the
signal is mixed with noise, and the received pulse amplitude is therefore
xo+t, where t is the noise amplitude (positive or negative). Suppose we
are interested in recovering the message with an accuracy of 0.001, i.e.,

the relative error is AA 0.001. To this end, we should have
0

1EI< 3 Axo. (3.6)

If || 2 const, i.e., the noise is constant, this condition is satisfied
when x; > 2000 Iĝ] or P =xi>4- 10%? = 4.108 P,. In other words, the
signal must be a factor of four million more powerful than the noise level
(66 dB). This is the threshold signal power for the PAM transmission
of the instantaneous value of x(t) with an error not exceeding 0.001.
This case of constant-noise communication is trivial: constant
noises are easily corrected. The main difficulty is that the real noise
is a random function which cannot be corrected. Random noise, in
general, may take on arbitrarily large values, although the probability
of this event is low. To determine the threshold power in the presence
of random noise, we have to find the probability that the noise does not

1
exceed 2 Mo i.e., the probability that condition (3.6) is satisfied. If

we are dealing with Gaussian noise, i.e., noise with normally distributed
amplitudes, the sought probability is

! zqp[-55m ahs
po (181 52) - (I7) - 0. (3.7)
and the probability of error is
p-1—-p,-21—9(2), (3.8)

where o is the parameter of the Gaussian distribution, o= VE'—- V P, and ®
is the Laplace function, or the probability integral. This integral has been
tabulated in detail, and the sought probability can be extracted from the
corresponding table. For Axo— 106, the probability of error is of the order

of 1075, and then it falls off rapidly as Ate increases. For most practical

problems, the reliability corresponding to an error probability of 10-8 is
quite sufficient. We can thus ensure reliable transmission (in the above

HI. RADIO COMMUNICATION WITH EXTRA 1ERRESTRIAL CIVILIZATIONS

sense) with signal reproducibility of “2 0.001 if x9 1000 Ax; —10*c and
P
pi-105.

We have considered the determination of threshold power in the simplest
case of message transmission by a single pulse. The results, however,
remain valid for any complex electrical signal x(t). In the general case, xo
is to be interpreted as the minimum signal amplitude (x2 = P nin).

Note that for a given dynamic range and given minimum signal power,
the minimum power is also well determined. As the mean power is
reduced, the communication becomes unreliable. Thus, besides the
minimum threshold power Pmin=aPn, we can also speak of the threshold
mean power ofthe signal. Later, when dealing with the transmission
of continuous functions by pulsed signals, we will show how to determine
the threshold mean power for certain types of signals (PCM with an
arbitrary code base). Now we will consider the spectral characteristics
of a signal.

Any periodic function x(!)of period T can be written as a sum of
harmonic vibrations of multiple frequencies (a Fourier expansion):

x (t) -2 Cy COS (yl + Pr). (3.9)

Each component (harmonic) of this expansion is a sinusoidal vibration of
frequency o, amplitude c, and phase gr. The frequencies of the
individual harmonics are integral multiples, and are related to the period

of the function by the equality og =k Am (k=1,2,3,...). The lowest

frequency is o -3, and this is also the difference between the frequencies

of any two successive harmonics. The values of c, and q, depend on the
form of the function x(/). The set of the coefficients c, form the amplitude
Spectrum, and g, the phase spectrum. Such a spectrum, consisting of
individual discrete values, is known as a line spectrum. As the period
increases, the spacing between the lines decreases, and in the limit for
T— o (i.e., a nonperiodic function), we obtain a continuous spectrum
(Figure 35). Mathematically, a continuous spectrum is expressed by

a Fourier integral.

Knowledge of the amplitude spectrum and the phase spectrum
completely defines the function x(/)J. Therefore, any process may be
described either by defining the appropriate time function or by specifying
the spectrum, which is a function of frequency. Both the time and the
frequency representations are equivalent.

Allthe signals encountered in practice are bounded-spectrum
functions. This means that they do not contain frequencies below some
minimum frequency v; and above some maximum frequency vz.

They occupy a finite frequency band from v, to v». The band of fre-
quencies filled by the spectrum of the signal defines the signal bard
width Av=v.—. This is a highly important characteristic of the signal.
In transmission along a communication channel, the signal frequency
band may shift toward higher or lower frequencies in the spectrum.

EXTRA TERRESTRIAL CIVILIZATIONS

However, the band width Av remains unchanged by this shift.* The frequency
Shift is very useful in radio engineering, e.g., in superheterodyne re-
ceivers. The application of this effect in communication systems makes
possible simultaneous transmission of numerous messages along a single
communication line, by using different frequencies.

The greater the band width Af of the communication line, the higher
is the number of signals with a given band width Av that can be transmitted
simultaneously. Each signal is associated with a certain message,
characterized by a definite quantity of information. We thus conclude
that the rate of information transmission through a
certain communication channel is proportional to the
channel band width.

Z w @ EA
T 7

FIGURE 35. The spectrum of a periodic pulse train.

The vertical axis gives S, e c&T (the product of the amplitude of the corresponding harmo-
nic and the period). The dashed line gives the spectral density of the amplitude of a unit
pulse. As the period is increased, the spacing between the spectral lines diminishes and in
the limit 7 -> œ a continuous spectrum is obtained, which coincides with the spectrum of a
unit pulse.

* In certain stages of the transmission process, the signal band width Av may indeed change. Thus, in FM,
the band width of the signal in the communication line is n times greater than the band width of the
modulating function (n is the frequency modulation index). However. after demodulation, the receiver
reconstitutes a signal with a band width Av corresponding to the band width of the modulating function x(t).

Ill, RADIO COMMUNICATION WITH EXTRA TERRESTRIAL CIVILIZATIONS

Relation of pulse length to pulse band width. Number of
pulses transmitted through a channel of given band width Aj

A basic relation exists between the pulse length and the band width of
the pulse spectrum Av:

tAv = const. (3.10)

It follows from this relation that the band width of a pulse is inversely
proportional to pulse length. The numerical value of the constant depends
on the shape of the pulse. In all cases, however, this constant is of the
order of unity, and for some pulses (e.g., square pulses) it may even be
taken equal to unity.

Equation (3.10) is a very general relation which is valid for any time-
variable process of duration t. Hence it follows that a continuous time
function x(t) with a band width Av and duration Af2Av'is of necessity a
combination of several individual pulses of various durations t;«A!, the

shortest of which is of duration t of the order of xe

Let us now determine the number of pulses that can be transmitted in

p|-
S.

unit time through a channel of band width Af. Let the pulse duration bet,
The band width of this pulse is Av, --A (we took the constant in (3.10)

to be equal to 1). Since the channel band width is equal to the pulse band
width, allthe frequency components of the pulse will be transmitted

through the channel and the pulse will be reconstituted without distortion
at the receiving end. The total number of pulses transmitted through the

channel in unit time is = =Af. Now suppose that the pulse is 10 times

longer, t= $ . The band width of this pulse is 1/10 of the band width of
the previous pulse, Av) uL 0.1 Af. Separating the signals in frequency,
we can accommodate in our communication line 10 frequency channels of
width Afeach. Each of these channels will transmit n0 Af pulses in

unit time, and the total number of pulses transmitted through all the 10
frequency channels will be Af as before. Finally, let the pulse duration

be VES. The band width of each pulse is then greater than Af. The

pulse components with frequencies v A/ are not transmitted through the
communication channel, and the signal is distorted. It may therefore
seem that Af determines the maximum number of pulses which are
transmitted without distortion in 1 sec through the communication channel.
However, this is not exactly so; a more rigorous treatment shows
that the maximum number of pulses is double this quantity, being equal
to 2Af.

Indeed, let an ideal frequency filter with a pass band Af be mounted
at the entrance to the communication line. At the time /— O0, a brief pulse

(s <5) of arbitrary shape is delivered to the filter input. After passing

through the filter, the pulse becomes blurred and its shape is described by
the function

EXTRATERRESTRIAL CIVILIZATIONS

x (t) = xg

SCANS (3.11)

where xo is the amplitude of the original brief pulse. The properties of
this function are responsible for the fact that the communication channel
is capable of transmitting every second a number of pulses equal to double

FIGURE 36. Illustrating the determina -
tion of the number of pulses transmitted
in unit time through a channel of given
band width. After transmission of a
short pulse of arbitrary shape through
an ideal low-frequency filter with a
pass band af, the pulse is distorted to
the shape shown in this figure:

sin 21 Aft
D x= n- AM
1
sin 21 Al (t s;
2) x(t) =X) ————;- 774
23 ar(t- rs)
2
sin 21 Af (t ——
3) x)= poen

3
22 at (17 zr)

The curves correspond to different
pulses with amplitude x». x. x:
which are delivered to the filter input

at the times t=0, E t= TAF

At the time t;, the amplitude of the
i-th pulse (after transmission through
the filter) is x,, and the amplitudes

of all the other pulses are zero. The
combination signal at the time t; is
therefore entirely determined by the

amplitude x; of the initial signal.

the channel band width. Function (3.11) is
shown graphically in Figure 36. For /—0,
x—xXo; for t=!/Af, *hAf, *hAF, .. ., x(t) = 0.
If we now send a train of brief pulses at
equal time intervals Af- '5Af, we obtain
some combination signal, a sum of signals
of the form (3.11) displaced by an amount
iAt (i=1,2,3...) relative to /=0. This
combination signal has the form

sin 2x Af pokes
iow Sie ent gl
i

Since each term of this sum is equal to zero
at any of the times t;=jA/ for j=1,2,3...
except j=i (see Figure 35), the combination
signal at any of the sending times t; is
determined only by the amplitude x; of the
corresponding brief pulse. Thus, despite
the distortion of short pulses after trans-
mission through a filter of band width 4f,
these pulses following one another at a rate
of 2Af pulses per second will be fully re-
constituted if the pulses at the receiving

end of the line are measured at the same
rate (at intervals At—'/;Af).

(3.12)

Transmission of continuous functions
by pulsed signals

A continuous message function x(t) of
duration A7=f,—t, can be represented by
a sequence of discrete values x(/,)taken
at time intervals A4. The representation
is clearly of higher accuracy for small
time intervals A4. The discrete values
of the function can be transmitted through
the communication channel using one of
the pulse modulation systems. If Af is the
channel band width, the maximum number
of pulses than can be transmitted in unit

time through this channel is 2Af. Using a succession of pulses following
one another at this rate, we obtain at the receiving end a time function
z(t) which is expressed by (3.12). In a noise-free channel, the values of
this function at the quantization times & are determined entirely by the

Ill, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

values of the original function, %(f,)=x(t,). The question is, are these
functions equal at any time /, and not only at f, i.e., are they identically
equal? The fit between the two functions is naturally improved if the
original function varies slowly between the quantization times fr. This
means that the function should not contain very high harmonics. According
to Kotel'nikov's theorem, the two functions are identical if the original
function x(f)does not contain components with frequencies v higher than
Af, i.e., if the band width of the Av transmitted function is equal to the band
width of the communication channel. Kotel'nikov's theorem is highly
significant for the theory and technology of communication, since it
permits converting continuous functions into a train of some discrete
magnitudes for transmission. This theory maintains that a function with

a bounded spectrum Av is completely determined by its values measured

at intervals At='/2Av. In particular, a function of duration A/, i.e., a
function which does not vanish only for tph<t<fp+At, is determined by a set
of 2AtAf discrete values. Thus, the definition of information derived for
discrete messages can be safely applied to continuous functions with a
bounded spectrum.

When continuous functions are transmitted by means of pulsed signals,
the main difficulty is that the function may take on any instantaneous
values, including irrational and transcendental numbers with an infinite
number of significant digits. Theoretically (in a noise-free channel),
these numbers can be transmitted with full faithfulness by PAM or another
suitable technique. In reality, however, reconstitution of the original
pulse with sufficient accuracy (or transmission of a sufficiently high
number of significant digits) in a noisy channel requires an excessively
high signal-to-noise ratio in the communication channel. "Therefore, the
next Step adopted in the transmission of continuous functions calls for
quantization of the message. To quantize the message, we select from
among all the values of x(f)a set of N discrete allowed levels xj x; ... xx.
which are distant Ax from one another (the quantization gap). All the
other values are regarded as forbidden. Only the allowed values are
transmitted. If the true instantaneous value of the function falls inside
the interval (x; xi41), i.e., takes on a forbidden value, the nearest allowed
value, differing from the true value by less than half the quantization
gap, is transmitted through the channel. This operation is completely
analogous to the rounding -off of numbers; it essentially signifies that
we are transmitting the true values of the function up to a certain number
of significant digits.

The quantized values of the signal in the communication channel are
affected by random noise. The width of the quantization gap should be so
chosen that with a given probability p the noise does not exceed half the
quantization gap. Then the signal can be accurately reconstituted at the
receiving end of the channel, since in this case the signal level nearest
to the noise-distorted value is the same as that fed into the communication
channel. The probability of signal reconstitution error is equal to the given
value p. 'The reconstituted signal can be again sent through the communica-
tion line, and this procedure may berepeated severaltimes, without affecting

EXTRATERRESTRIAL CIVILIZATIONS

the reconstitution of the original quantized level. The transmission of
quantized values instead of the true values is equivalent to superimposing
a certain noise 6, which does not exceed half the quantization gap. This
noise is known as quantization noise. Quantization thus does not free the

m
ANNANLA YS

2
b
a
4
g
2
7

FIGURE 37. Transmission of messages by pulse code:
(a) a continuous message function x (t); (b) quantized
values of the function; (c) transmission of the quantized
function by binary code; (d) Baudot telegraph code;

(e) Morse code.

signal from noise, but in effect substitutes one kind of noise for another.
The random uncontrolled noise is replaced with an artificial noise — the

quantization noise. The intensity of this noise is not weaker than that of
the undesirable natural noise. However, the advantage of a quantized

IIl, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

System is that the noise is fully controllable and the accumulation of
random errors is avoided.

Another important advantage of quantization is that it permits trans-
mitting continuous message functions by means of pulse codes. Each
discrete value of the function is expressed by a certain positive number,
i.e., it can be written in any number system in the form of a certain
numerical sequence. The electrical signal corresponding to this discrete
value of the function similarly can be represented as a combination of
individual electrical pulses. The various pulse combinations correspond-
ing to the various values of the message function constitute a certain
code. Every individual combination is regarded as a code combination.
Various elementary signals (pulses) used to construct the code combination
are known as the code elements, and the number of different elements used
in the code combination is the base of the code.

The Morse telegraph code is an example of a ternary code. Its
elements are a short signal, "dot", a long signal, "dash", and the
absence of a signal, a blank of the same duration as the "dash" intended
to separate successive letters. The number of symbols in Morse
code combinations is variable; this is a nonuniform code. The Bodo
telegraph code is a five-digit binary code; its elements are a pulse and an
absence of a pulse, both of equal length. Each message (a letter of the
alphabet) is represented by a five-element code combination (Figure 37).

If a is the base of the code, m is the number of elements in a code
combination, the total number of code combinations or different values
that can be transmitted by this code is N=a™. Quantization makes it
possible to transmit functions using a code with a finite number of elements
m in each code combination. Without quantization, N=oo and in general an
infinite number of code elements in a code combination will be needed to
transmit the true instantaneous values of the function x(t).

The transmission of continuous functions by a pulse code whose
elements are pulses differing in their height h only is known as pulse-
code modulation.* In PCM transmission, the signal is first confined to a
limited band width, so that all the frequencies above a certain v, are cut off.
Signal readings are then taken at a rate of 2v per second. The readings
and quantized and encoded. 'The number of elements m in a code combina-
tion for a given base a is determined by the required number N of quantum
levels. Thus, in telephone communications, the best sound is achieved for
N=100. Therefore, a seven-digit binary code can be used for the PCM
transmission of telephone conversation (27 = 128). Code groups are
delivered to the communication line. At the receiving end, the pulses
distorted by noise are reconstituted, the code groups are decoded, and a
new sequence of pulses with amplitudes proportional to the initial quantum
values xj xe ..., xy is formed. These pulses, coming at a rate of 2vo pulses
per second, are transmitted through a low-frequency filter with a cutoff
frequency vo, and are then combined to give the original signal.

* Generally, PCM is regarded as transmission of message by binary pulse code. This technique is often used
in practice. However, theoretically, we may consider PCM for any code base.

EXTRATERRESTRIAL CIVILIZATIONS

Let us find the threshold power of PCM. (Threshold power in this case
is defined as the threshold mean power, rather than Pg.) Consider a code
with a general base a; let Ah be the difference in the pulse heights cor-
responding to two successive elements of this code. The power of the i-th
pulse is P,=x?=(iAh), and the mean signal power, assuming a uniform

frequency of occurrence of all the pulses, isP,— ŽEP.. This power is

minimum íf both positive and negative pulses are used to make it up. Then,

a-l

2
(AA)? 2. (Ah)?
p, = Ae $ Pen e (3.13)
a
PEN

To ensure correct reconstitution of noise-distorted signals, the random
noise § should not exceed half the value of A^ (Itl « z^). The probability
of this even, as we have seen before, depends on the ratio M, For Ah=100

the probability of an error (i.e., an incorrect reconstitution of the pulse)
is 1078, Inserting this value of Ah in (3.13), we obtain the threshold power
P; (a) for a PCM system with a code of base a:

P? (a) = 1g 0? (at — 1) = 299 (a — 1) p,. (3.14)
For a given noise power, the threshold signal power increases with the
increase of code base. The maximum threshold power is observed for
a=N, i.e., for ordinary PAM. The threshold power of PAM is

Pran = (N? — 1) P, = MO p, (3.15)

The threshold power in this case is seen to be proportional to N*. For

any other code base a#WN, the threshold power is independent of the number
of quantization levels N and is determined by the code base only. Fora
given a, there should be m pulses in each code group to encode the quantum
levels (since N=a™), If we reduce the base a, mis increased correspond-
ingly, i. e., the number of pulses transmitted through the line in unit time
increases. PCM thus enables us to reduce the threshold signal power by
increasing the band width of the communication line. The minimum
threshold power is attained when using binary code. In this case P? (2) =25P,.

Transmission rate of a communication channel

We have now reached the stage when the transmission rate of a
communication channel can be determined. On p. 76 we mentioned
that the transmission rate is proportional to the channel band width. The
signal-to-noise ratio also plays an important part in this respect.
Consider a message which constitutes a table of three-digit decimal
numbers. We have a channel of 3 kHz band width and a signal-to-noise

ratio fs -25. Using (3.14), we find that for this signal-to-noise ratio
n

IH. RADIO COMMUNICA TION WITH EXTRATERRESTRIAL CIVILIZA TIONS

the code base is a=2, and from the relation N=103 =a" we find m=10, i.e.,
a ten digit binary code can be used to transmit the message through our
channel. A channel with 3 kHz band width will transmit 6000 pulses per
second, or 600 code groups of 10 bits each. The quantity of information
contained in each code group is Q,— 310g; 10 = 10 log, 2= 10 bits. The
transmission rate of the channel is therefore 6000 bits per second. Now
suppose that the transmitter power is increased by a factor of 5, so that the

signal-to-noise ratio becomes i =125, If we are using binary code, as
n

before, the channel transmission rate for the given band width (3 kHz) natu-
rally does not change. However, the binary code is not very efficient

for such a high signal-to-noise ratio. The transmission rate can be raised
by using a different code system. From (3.14) we find that for

Ps
Pn
may thus use a five-digit quaternary code. Transmitting as before 6000
pulses per second, we may now transmit 1200 code groups of five quaternary
pulses each. The quantity of information associated with each code group

is 10 bits as before (3 log, 10=5 log,4= 10) and the transmission rate is

=125, we may take a=4. Now from N=108=am, we get m=5. We

therefore 10X 1200 = 12,000 bits per second. For He 825, we may use a
n

three-digit decimal code, raising the transmission rate to 10x 2000 =

= 20,000 bits/sec. Finally for a signal-to-noise ratio equal to 8-109, we
may take a= 10°=N, m=1, i.e., transmit using the ordinary PAM (without
coding). Each pulse corresponds to a three-digit decimal number and thus
contains 10 bits of information. A channel of 3 kHz band width may
transmit 6000 such pulses and the transmission rate of the channel will
therefore be 6-104 bits per second. The same quantity of information can
be transmitted for Fi — 25, using a binary code and increasing the channel
band width from 3 to 30 kHz.

This example clearly illustrates the importance of each factor affecting
the channel transmission rate. The frequency band determines the number
of pulses that can be transmitted through the channel in unit time. The
signal-to-noise ratio gives the base of the code that may be used for
transmission through the particular channel and, hence, the information
content of each pulse. Thus, in binary code transmission, each pulse
carries 1 bit of information, with ternary code each pulse carries 1.6 bits,
in quaternary code 2 bits, in decimal code 3.3 bits, etc. By reducing
the code base, we lower threshold power of the system and at the same
time lower the quantity of information carried by each signal, so that to
ensure a constant transmission rate the band width must be increased.

Let us find the transmission rate of a PCM channel. Let the band
width of the communication line be Af. Then it will carry 2Af=nm
pulses per second, where n is the number of code groups transmitted
each second through the communication channel, and m is the number
of pulses in the code group. The information Q, associated with the
transmission of each code group is Qi 711ogs a, and the total quantity of
information transmitted through the channel in 1 sec is

q = nQ, =nm log, a —2Aflogsa = Af log, a?. (3.16)

EXTRATERRESTRIAL CIVILIZA TIONS

Inserting a? from (3.14), we obtain

12 He)
9=Aflog,| 1 +22 22) . (3.17)

This is the maximum transmission rate of a PCM system. If the code

base a is chosen so that P? (a) is the mean signal power in the communication
line, the P? (a) in (3.17) can be replaced by P,. For any other code base b
(2<b<a),

2af «qal log (Ls pt). (3.18)

A useful characteristic of a communication system is the ratio 3"

which characterizes the transmission rate per 1 Hz. 'The corresponding
values for PCM are listed in Table 3.1.

TABLE 3.1. PCM transmission rate per unit band width

Number of bits per 1 Hz

Code base, a Threshold power P,/P, Number of bits per pulse q T ( 12 2a)
"Apo 8

2° 25 1.0 2.0
3 61 1.6 8.2
4 125 2.0 4.0
5 200 2.8 4.6
6 292 2.6 5.2
1 400 2.8 5.6
8 525 3.0 6.0
9 666 3.2 6.4
10 825 3.3 6.6

The PCM coding system is not optimal. The transmission rate
expressed by (3.17) therefore does not realize the full potential of the
communication system. Shannon /5/ has shown that there exists some
coding system, which in general may be quite complex, for which the
transmission rate can be raised to

q= Af log (1 +>). (3.19)

This coding system is termed ideal.

Shannon's equation (3.19) described the maximum transmission rate of
a channel of given band width Af and given signal-to-noise ratio x No
communication system, however complex and sophisticated, will transmit
information at a higher rate for the same Af and $i. Shannon's formula thus

establishes the limiting relation between the basic parameters of a com-
munication system, systems of communication with extraterrestrial
civilizations included.

IH. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

Figure 38 plots the rate of information
transmission per 1 Hz, F as a function of

the signal-to-noise ratio in a communication
channel for an ideal Shannon system and

for PCM. The ideal coding system ensures
a gain of 8—10 dB in power compared to

the PCM. Moreover, the PCM has a sharp
threshold power P? (a), determined by the
assumed error probability. Both the
threshold power and the associated numeri-

cal coefficient before * in (3.17) change

Bits per sec/Hz

-20 -0 0 W 2
Signal/noise ratio, dB

when the error probability is changed.
For P,«P?(a), information cannot be

FIGURE 38. The transmission rate of a transmitted with the specified reliability
communication channel. (the specified error frequency). An ideal
The solid curve corresponds to Shannon's system does not have a clearcut threshold
ideal system. The dots refer to PCM power. It may operate for any P,,

with positive and negative pulses for ensuring reliable transmission of informa-
error frequency of 1075; the numerals tion according to (3.19) with any arbitrarily
next to the PCM dots correspond to the small error probability. In particular, for
code base. P,

p ^ 3, g=2Af, i.e., the ideal system
has a transmission rate equal to the transmission rate of a binary PCM

system (for a threshold signal-to-noise ratio $ - 25) For E =1,
n

qg=Af for the ideal Shannon system, and then it rapidly decreases,
reaching zero for F =0. Finally, for P, — 0, B — oo and g also goes
to infinity, i.e., the rate of information transmission through a noise-free
channel can be made arbitrarily large. This is also true for PCM. In
practice, this feature can be realized in PCM systems by using a code witha
very large base a. Indeed, any text may be represented as a number with
sufficiently numerous significant digits. 'This number can be transmitted
through a noise-free channel as a pulse of appropriate height.

Let us consider the dependence of the maximum transmission rate of
a channel on band width. The P, entering Shannon's formula (3.19) depends
on the band width. In most practical cases, we may take

Py = Pra. gp Sf (3.20)

Here P,,,, is the noise power per unit frequency interval, called the
specific noise power. Inserting P, from (3.20) in (3.19), we find

q=Aflogs(1 + o). (3.21)

EXTRATERRESTRIAL CIVILIZATIONS

If we take Af, = 3. i.e., define Afg as the band width for which the noise
n.sp

power is equal to the signal power, we may write (3.21) in the form

3 loge (1 + A). (3.22)

am o Af

Figure 39 shows -+ EU as a function of rA

As the band width increases, the transmission rate rapidly grows up
to a point where the signal power becomes comparable to the noise power
(for Af=Af,). After that point, the growth of the transmission rate is slowed
down, and for Af — oo, it goes asymptotically to the transmission rate for

Af=Afo multiplied by logge = 1.443.

dd EM 1443
- EM

Bits per sec/Hz

7 2 I 4 A4

FIGURE 39. The transmission rate of a communica-
tion channel as a function of the band width. Af,
is the band width for which P, =P, .

$2. RANGE AND INFORMATION CONTENT OF INTER-
STELLAR COMMUNICATION

The optimum communication frequencies

We have considered some applications of the general theory of com-
munication, and now we can proceed with a discussion of the problem of
communication with extraterrestrial civilizations. The main difficulty
of setting up a system of communication with extraterrestrial civilizations
is that different elements of the system belong to different "subscribers,"
and we have no advance knowledge of the type of instruments they are using.
As a result, every subscriber, whether on a transmitting or a receiving
end, should see to it that the signal transmission and reception devices
ensure reliable radio communication despite this intrinsic uncertainty.

In this general formulation, the problem includes the various aspects of
coding, call signals, signal detection (including the criteria of artificial
origin of signals), and signal decoding. Some of these topics are considered

elsewhere in the book.

Ill, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

A schematic diagram of an interstellar radio-communication system
is shown in Figure 40. The message from the sender (EC-1) is delivered
to a transmitter, which converts it into a signal, and the signal is then
radiated into the outer space by the transmitting antenna /l,. At the
receiving end of the conimunication line, the radio waves are picked
up by a receiver antenna A, and the electric signal is directed to the
receiver where, after various transformations, the original message
is reconstituted.

[LT a
Message S 14 4i 2
Trans- | Signal ! . , l Recei
Ec] [5] Z <7 Line AT ver [66:3
!
bossa Aa one

FIGURE 40. A diagram of a sys'em for interstellar radio communication.

A, ~ transmitting antenna, .1: — receiving antenna, /.— transmitter
power, P — antenna radiation power, 7, — special flux density at observa -
tion point, P, — signal power at receiver input.

The communication line is the common element of the system joining
the two "subscribers." In interstellar radio communication, the line
comprises the part of the outer space between the transmitting and the
receiving antennas (the interstellar medium plus the corresponding planetary
atmospheres) where the radio waves propagate. The line parameters
depend on the conscious activity of the "subscribers," as well as on certain
objective factors, such as radio wave absorption in the interstellar medium.
We have seen in Chapter II that the absorption coefficient of the interstellar
medium increases with the decrease in frequency. Over large distances
(of the order of the galactic diameter), the interstellar medium is virtually
opaque at meter wavelengths. This automatically limits the range of wave-
lengths for interstellar communication: because of strong absorp-
tion, interstellar communication is unfeasible at fre-
quencies shorter than 1 MHz.

Another important objective factor is the noise in the communication
line. The various noises can be dívided into two groups: instrumental
noise and background noise. Instrumental noise is controllable and it can
be reduced to a comfortably low level. Background noise is determined
by the radio emission of the planetary atmospheres and the radio waves
originating in the outer space. Atmospheric noise in principle can be
eliminated by mounting the antennas at an appropriate distance from the
planetary surface, e.g., on artificial satellites. Noise associated with
radio waves from space is intrinsically unavoidable.

Another source of intrinscially unavoidable noise are the quantum
fluctuations,* associated with the quantum nature of the electromagnetic
radiation.

* Not to be confused with quantization noise ($1).

EXTRATERRESTRIAL CIVILIZATIONS

Background noise and quantum fluctuations determine the optimum
frequency range of electromagnetic waves for interstellar communica-
tion. This problem was analyzed in some detail in Chapter I. We have
seen that the optimum frequency range for the purposes of
the search for call signals of extraterrestrial civiliza-
tions is confined to the region of minimum background
noise (A= 10—50cm), and for reception of meaningful
messages to the region of minimum sky brightness
temperatures. The last condition is satisfied for a very
wide range of frequencies, from decimeter to sub-
millimeter waves.

The choice of the exact working frequency band in the optimum fre-
quency range requires a separate discussion. This topic was also
analyzed in Chapter I, where we derived an expression for the optimum
distribution of the transmitter energy in the spectrum, needed to ensure
maximum information transmission rate. For moderate quantities of
information, the question of the transmission rate is not particularly
acute, and the frequency band may be taken fairly narrow. In this case,
we are faced with the problem of frequency scanning in our search for
signals. Cocconi and Morrison /1/ proposed using the frequency of the
hydrogen radio line at 21 cm (v= 1420 MHz) or one of its harmonics.
Similarly, the frequency of the hydroxyl OH radio line at 18 cm can be
used. Troitskii /6/ suggested that the search should be conducted near
the radio lines of individual molecules used in masers (the 1.25 cm
ammonia line and the 0.4 cm formaldehyde line).

Range of communication

An important parameter of a communication line is its length or extent.
Since to first approximation we may assume that the civilizations are
uniformly distributed in space, the number of probable subscribers and,
hence, the probability of establishing communication is proportional to the
cube of the communication range. What factors determine the communica-
tion range? The first step is to define exactly the concept of communication
range. We are dealing with two problems: detection of EC signals and
reception of meaningful messages. Accordingly, we will discuss the
range of detection and the range of communication for the
reception of meaningful messages. Before any meaningful information
can be received, we have to detect the EC signals. However, the expres-
sion for the range of communication is simpler to derive, and we will
therefore start with this concept.

'The range of communication is equal to the maximum distance over which
the communication system is capable of transmitting and receiving informa-
tion with a given reliability (a given error probability). Over greater
distances, the signal power falls below the threshold value, and the signal
cannot be reconstituted with the required reliability.

Let us now derive an expression for the communication range. Let Po
be the power of the EC-1 transmitter, Aj, the frequency band of the trans-
mitter, n the efficiency of the transmitting antenna. The power radiated
by the antenna is then P;=nPy. If this power is radiated isotropically,

II, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

i.e., uniformly in all directions, the radio flux at the observation point
at a distance R is

F,Af = oh. (3.23)

Here F, is the spectral energy flux density, or the energy flux per unit
frequency band.

Realantennas arenot ideally isotropic: they have certain directional
properties. The directional properties of the antenna are characterized
by its directivity pattern or diagram. The directivity pattern of the
transmitting antenna is a polar diagram which plots the energy flux
radiated by the antenna in various directions. Figure 41 shows the
directivity pattern of a reflector antenna. Almost the entire energy is
radiated by this antenna within a certain small solid angle accommodated
by the main lobe of the pattern. If the antenna has a rectangular cross
section with sides / and A, the angular width of the main lobe in the
corresponding directions is

20,=24 and 20, —2 *, (3.24)

where 4 is the wavelength. For a circular reflector antenna (e.g., a
paraboloid of revolution), the width of the main lobe is

20,—2 x 1.225 «1.223, (3.25)

where r is the radius, D is the reflector diameter. This quantity is
usually referred to as the width of the antenna pattern or the beam width
at zero power level. Another significant parameter is the beam width
between half-power points, which for a circular cross section antenna is
expressed by the equality

A A
20,,—2 X 0.515 =>. (3.26)
To first approximation, the antenna pattern may be regarded as constant
(equal to its maximum value) within the beam width angle, falling to zero
outside this angle.

The plane Æ

FIGURE 41. The directivity pattern of an antenna.

EXTRATERRESTRIAL CIVILIZATIONS

In calculations of the radiated power, we will use the directivity co-
efficient of the antenna. The directivity coefficient of a transmitting
antenna is equal to the ratio of the antenna power radiated in a certain
direction (e.g., along the axis) in a unit solid angle to the mean power
radiated in a unit solid angle in all directions. In other words, the
directivity coefficient is defined as the ratio of the energy flux radiated
by the antenna inside a small angle do to the energy flux radiated by an
isotropic radiator of the same power in the same solid angle de. When
using a directional antenna with a directivity coefficient g,, the radio
flux at the observation point at a distance R will be

Fy Af = E, = D = Pete (3.27)
The quantity e-—n£giis known as the antenna gain. If the transmitter power
P, and the antenna gain are known, the radio flux can be determined without
difficulty at any observation point. In what follows, we will assume for
simplicity N= 1,P,— Ps, £17— £t.

Receiving antennas are also directional. In the theory of antennas it
is proved that, in virtue of the reciprocity principle, the antenna
properties are the same in transmission and reception. In particular,
the antenna pattern, the directivity coefficient, and the gain of the receiving
antenna are equal to those of the same antenna working as a transmitting
antenna (when a transmitter is connected to the antenna terminals).

The power P delivered by the antenna to the receiver is clearly
proportional to the radio flux at the reception point. We may therefore

write

P,—SF,N. (3.28)

S, expressed in cm?, is the effective area of the receiving antenna.

This quantity is equivalent to the exit aperture of an optical telescope.

In particular, for a reflector antenna with n=1, the effective area is
equal to the geometrical area of the reflector. The effective area and the

antenna gain are related by the equality
Sc (3.29)

We can now derive an expression for the range of communication as a
function of the parameters of the transmitting and the receiving systems.
The signal power P, at the receiver input substantially depends on the ratio
of the transmitter to receiver band width. Two possibilities should be
considered here.

a) The receiver band width is greater than the trans-
mitter band width (A-Af).

This case is observed, e.g., for the reception of narrow-band mono-
chromatic signals. Using (3.27) and (3.28) and introducing the subscript 1
to identify the parameters of the transmitting system and subscript 2 to
identify those of the receiving system, we find

P, = SFs Af = SE (3.30)

5780 90

HI. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

Note that the result is independent of the receiver band width ^/;and, for
a given transmitter power P;, it does not depend on the transmitter band
width either. The noise power at the receiver input, as we know, is
proportional to the receiver band width:

Pa = Py. sp Afe m RT, Afo- (3.31)

Here k is Boltzmann's constant, equal to 1.38: 1016 erg/deg, T, is the
noise temperature, generally introduced as a parameter of the noise power.
It is equal to the temperature of an active load (a resistor) matched to the
receiver input which produces the same noise power when connected in
place of the antenna. When dealing with background noise, 7, is the
equivalent brightness temperature of the noise radiation. In particular,
if the background is associated with the thermal radio emission of some
space medium, 7, coincides with the temperature of that medium.

The last two expressions give the signal-to-noise ratio at the
receiver input:

P Pigs.
a= pi TARET My (3.22)
In $1 we saw that this ratio describes the reliability of communication.
For reliable communication, a moreover should exceed a certain threshold
value, which depends on the particular coding system used. In usual
communication systems, a » 1. Equation (3.32) shows that the re-
liability of interstellar radio communication is propor-
tional to the transmitter power multiplied by the
transmitting antenna gain and the effective area of the
receiving antenna and is inversely proportional to the
noise temperature, the receiver band width, and the
Square of the distance between the civilizations.

For givena, the distance R at which the required signal-to-noise ratio
is attained can be found from (3.32):

zi Pigis !h
R= (sis A) Ld

or, using (3.29),

PISIS. Mh
R - (air: ss) ' (3.33b)

Pigg:

[A
R= (jetta) > (3.33c)

i.e., the range of radio communication increases with the
increase in the transmitter power and the directivity or
the effective area of the receiving and the transmitting
antennas; it also increases with the decrease in noise
temperature and the receiver band width. The dependence on
à in (3.33b) is attributed to the fact that, for a given area S; of the
transmitting antenna, the directivity increases at shorter wavelengths;

EXTRATERRESTRIAL CIVILIZATIONS

the dependence on 4 in (3.33c) is associated with the fact that, for a given
ge, the effective area of the receiving antenna increases with the increase
in wavelength.

b) Let us consider the second case: the transmitter band width
is greater than the receiver band width (Af;i>Af). This case
is observed for the reception of wide-band signals, e.g., when the
transmitter energy distribution is determined by the requirement of
maximum information content (see Chapter I). The spectrum of the
Signal in this case is limited by the receiver band width, and the receiver
P, is given by ;

P, = SF, Af, = Eh, (3.34)

i.e., in distinction from case a, the signal power is proportional to the
receiver band width Af,, and for a given total transmitter power, it is
inversely proportional to the transmitter band width. The noise power
is expressed by (3.31), as before, so that the signal-to-noise ratio (for
a given range) and the communication range (for a given signal-to-noise
ratio) are respectively given by

Ps _ pns

P, = TARE ANRT,’ (3.35)
XN P g S Ma

R = (xcu. ) i (3.36)

Comparison of these expressions with (3.32) and (3.33) shows that
they differ only in the subscripts of Af. In the former case, the signal-
to-noise ratio and the range of communication increase with decreasing
receiver band width and, for a given transmitter power, are independent
of the transmitter band width. In the latter case, conversely, the signal-
to-noise ratio and the resulting range of communication increase with
the decreasing transmitter band width and are independent of the
receiver band width. In general, we may thus write

ac AF, Ro AFP, (3.37)

where Af is the greater of the two band widths Af, and Af.

Let us consider the range of communication as a function of the
parameters of the transmitting and the receiving antennas. This dependence
is expressed by (3.33), where Aj, should be replaced with Af=max(Af;, Af;).
Setting g:=g.=1 in (3.33c), we obtain the range for the case of isotropic
transmission and nondirectional reception. Taking g.=1, we obtain the
range for directional transmission and nondirectional reception. Finally,
taking g,=1 in (3.33a), we obtain the range for isotropic transmission
and reception with a directional antenna of effective area S;.

Let us consider the dependence of ~ and Ron band width. Equations
(3.32) and (3.33) are conveniently written in logarithmic form:

1 1 S. 1
lg R= y le Pig + 518 araar, — y IgA. (3.38)

IIl. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

The gain (attenuation) is generally expressed in decibels. The product Pigi
can be expressed in dB: W. Let P,g,2100 dB- W. This means that a 1 W
transmitter is coupled to an antenna with a 100 dB gain, or alternatively

a 1l kW transmitter is coupled to an antenna with 70 dB gain, or finally

a 1 MW transmitter is coupled to an antenna of 40 dB gain, etc. If the
product P,g,in (3.38) is expressed in dB- W, R in light years, S, in square
meters, the equation takes the form `

=Pg plig 3 Ligat-
lgR- +o le arr, Tlga- 16. (3.39)

Let Pigı=200 dB: W, S,=104 m?, T, = 10*K; then
1 1
IgR=6.4— > lga— Ig Af. (3.40)

Similarly, for the same parameters of the receiving and transmitting
systems, we obtain

Iga —12.8—21g R — Ig Af. (3.41)

Figure 42 plots the dependence of a and R on band width. For Af;-Af.
the receiver band width clearly should be reduced. The noise power at
the receiver input will also decrease, so that the effective power will not
change. Asa result, the signal-to-noise ratio at a given distance Ror
the range of communication for a given signal-to-noise ratio will increase.
This increase does not entail a loss of information content, since the band
width of the communication line (the factor determining the channel trans-
mission rate) in this case is limited by the transmitter band width Af.
Moreover, at a given distance, the transmission rate may be increased
by increasing the signal-to-noise ratio. The maximum range is attained
for Af,=Af,. Further decrease of the receiver band width is inadvisable,
since the noise and the effective signal will then increase to the same
extent. Moreover, further decrease of the receiver band width will
limit the transmission band width of the communication channel.

For Afo<Af,;, the signal entering the receiver is limited on the
frequency scale and distorted (e.g., in pulse modulation, the chopping
of the frequency band will cause blurring and interference of the pulses).
Moreover, the narrower band reduces the transmission rate of the
communication channel. If the full band width Af, of the line is utilized
at the transmitting end, a decrease of the frequency band width will
result in a partial loss of information. In general, some information
is also lost when the transmitter band width is only partly utilized,

Since the character of the signal and its time and frequency distributions
are not known in advance. Hence it follows that the case Af,<Af, is
unfavorable for communication. To avoid signal distortion

and loss of information in this case, we should increase the receiver band
width Af: to Af,. This increase of band width will not affect the range of
communication, since it increases both the effective signal and the noise.
However, for technical reasons, the receiver band width cannot be
increased indefinitely. Even special wide-band receivers can hardly be
expected to have a frequency band wider than 10% of the particular electro-
magnetic frequency used. Another technique of band matching calls for

EXTRATERRESTRIAL CIVILIZA TIONS

reducing the transmitter band width. This can be achieved in two ways:
a) without altering the transmitter specific power, i.e., the power per
unit frequency, and b) without altering the total transmitter power P.

g
8
87
s
oe
i5 i 7
e
EM ue
», A
8 s b
82 $4
7 "3 >
2 E TS

a t à 1 1
2723945674 272345487427
Log receiver band width 44 Log receiver band width 47
FIGURE 42a, Signal-to-noise ratio « vs. receiver band w idth Ar; for various transmitter band widths Af;.
The following system parameters were used; R = 100 light years, Pie, = 200 dB- W, $,7 104n?, Ta =10°K,
The band widths are expressed in Hz, For a given transmitter band width Af,, the signal-to-noise ratio
increases with the decrease in Af, until the equality Af,= Af, is attained, Further decrease of the receiver
band width Af, does not increase the signal-to-noise ratio a. When the band widths are equal. further increase
of the signal-to-noise ratio can be attained only by a simultaneous reduction of both the receiver and the
transmitter band widths,

FIGURE 42b. Communication range as a function of the receiver band width for various transmitter band
widths Aj, .

Pig, = 200 dB- W, s,— 10*m’, T, =10°K, The range is expressed in light years. In both figures the arrow
marks the band width for which the signal is equal to noise at a distance of 100 light years (using the given
cominunication parameters).

In case a, the total power and the flux at the observation point decrease
in proportion to the decrease in the band width, but the spectral density
remains unchanged. The fraction of the total flux or the fraction of the
transmitter power delivered to the receiver will thus increase, since the
signal-to-noise ratio and the range of communication are not affected. In
case b, the contraction of the frequency band entails a growth of the
specific transmitter power and the spectral flux density F, at the observation
point. The total flux F,Af; remains unchanged, but the fraction of the total
flux intercepted by the receiver increases with the decrease in Afi As
a result, the contraction of the frequency band will increase the signal-
to-noise ratio and the communication range, as we see from (3.35) and
(3.36). The maximum range, as before, is attained for Af,- Af. Further
contraction of the transmitter band width is inadvisable, since the entire
flux at the observation point is anyhow intercepted by the receiver and,
for a given transmitter power P,, the signal-to-noise ratio remains
unchanged. If, however, the frequency band is reduced without retaining
a constant specific transmitter power, both the signal-to-noise ratio
and the communication range will decrease when it falls below Af;.

Thus the maximum range of communication is attained
for AlL-A5É. Once the bands have been made equal (either
by reducing the receiver band width forAf,>Af,, or by

II. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

reducing the transmitter band width forAfAAfP) the
communication range can be increased further only by
a simultaneous reduction of the transmitter and the
receiver band widths. This band width reduction will naturally
lead to loss of information. The channel transmission rate for a given
Separation R between the subscribers will decrease despite the increase
in the signal-to-noise ratio, since the dependence of the transmission rate
on band width is definitely stronger.

Range of detection

For ordinary systems, the range of communication is limited by
the condition «»1.* If, however, we are concerned merely with the
detection of signals, without decoding the information that these signals
carry, this condition is not necessary. The modern radiometric
techniques make it possible to detect signals which are much weaker
than noise. This is a common practice in radio astronomy, which deals
with extremely weak fluxes from sources in outer space.

The possibility of detecting such signals is based on the statistical
properties of noise. Had the noise power been constant, i.e., without
any fluctuations in time, it could have been easily corrected by intro-
ducing an appropriate voltage in the source, equal in magnitude to the
noise voltage and having opposite polarity. In principle, we could thus
measure signals of arbitrarily small power level. The measurement
procedure reduces to the recording of the small increment above the
constant noise level associated with the reception of the effective signal.
Correction, strictly speaking, is not absolutely essential: it only
constitutes one of the more convenient measurement methods.

Real noise, however, is a random process with voltage (or current)
amplitudes fluctuating at random about the zero level. If Af;is the
receiver band width, the mean duration of a single noise pulse Af, (or
the time during which the amplitude of the damped oscillations generated
by this fluctuation remains constant) is of the order of 3r. The number

of independent noise pulses observed in a time tz is thus n Tan Af,.
2

A recorder with a time constant t averages these noise pulses, and the
n
mean noise power P, a -i YP fluctuates (so-called recorder fluctuations)

about the theoretical mean noise power Pa, which is the expectation value
of the random power values P; of the individual noise pulses. To detect

a useful signal, the resulting noise power increment AP, =P, should exceed
the root-mean-square deviation of P, av from the theoretical value P,,

i.e., we should have P,=AP, >o,,(P)or

P, = Bo, (P), (3.42)

* For Shannon's ideal system, this restriction is of no consequence. Some special communication systems

also use incthods which permit reception of messages although the signal is much weaker than the noise.

EXTRATERRESTRIAL CIVILIZATIONS

where B is some dimensionless number greater than unity (po1). If
we are dealing with Gaussian noise, with a normal amplitude distribution,

we may write

9..(P) 1 —— 1 (3.43)

Py Vn VuM C

whence

P, Oay(P) i
-— epu 3.44
* P, P Pa V T: ^ ( )

This expression determines the minimum signal that can be recorded
with a radiometer. For f= 1, we obtain the limiting or the theoretical
radiometer sensitivity. The actual sensitivity is generally much lower,
since for reliable signal recording, f should be greater than 10. The
factor V-t Afa is known as the radiometer gain. For Yt,Af,>>1, the
signal-to-noise ratio may be much less than unity. For example, for
t= 1 sec and Af;— 10 MHz, the radiometer gain is 104; if B=10, we have
= — 107, i.e., the signal power is one thousandth of the noise power,
and yet it is 10 times stronger than the rms noise fluctuations and can be
reliably detected.

The detectability of signals which are weak compared to noise is
associated with the averaging action of the recorder, which averages
the individual noise pulses over a period of time equal to its time
constant. The effective signal is naturally also averaged, so that the
final result is the signal power averaged over the time t;. If the cha-
racteristic modulation time t is less than the time constant v of the
recorder, all the measurements related to signal modulation are smoothed
out and the information contained in the signal is completely lost. In
this case, we can only identify the presence of some effective signal of
mean power P,. It is in this sense that we will interpret the term "range
of detection," as distinct from the range of communication.

Reception of information requires that t:2t2. Using the relation
between the time and the band width, we rewrite this inequality in the
form*

Af, > nf, (3.45)

where n is the number of independent noise pulses averaged by the
recorder. We have noted before that for purposes of information reception
the receiver band width should not be less than the transmitter band.

* Here we take t, a i.e. use a coding technique which for a given transmitter band width Af,
1

ensures the maximum transmission rate. In general, when the sender does not fully utilize the
transmitter frequency band (intentionally Iowering the transmission rate of the communication
channel, to ensure a higher reliability at a fixed range and a higher range for fixed reliability),

T. " ; I ,
the characteristic modulation time t, can be greater than =>, and condition (3.45) is not satisfied.
1

IH. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

Condition (3.45) is stronger than that. It shows that in case of averaging,
it is no longer sufficient to ensure a receiver band larger than the trans-
mitter band. The receiver band should be greater than the transmitter
band multiplied by the square of the radiometric gain.

What is the actual range of detection? Inserting a from (3.44) into
(3.33) and (3.36), we obtain the following expressions for the range of
detection:

1) Af>Af,
— ( Paso Y We
R= (rt) (3) , (3.46)
2) Afe<Ah,
R= (aegis) (v, Af)" (3.47)

In the above example, when ~=1 sec, Af,210?Hz, the range of detection
can be increased by a factor of 100 due to the radiometric gain. For
Afe>Af;, the range slowly decreases as the receiver band is made wider;

for Afe<Af,, it also slowly (x Afi) increases, so that it is advisable to
increase the receiver band in this case. The maximum range is attained

for Afz=Afı, as before. In the absence of radiometric gain (Vx, 45; — 1),
equations (3.46) and (3.47) reduce to (3.33) and (3.36), as could have been
expected.

Let Af; be the given receiver band width. Consider two signals: a
narrow band signal with Afinar «Af;, and a wide band signal with Afi wide >Af2.
Let R, be the range of detection of the narrow-band signal and R, the range
of detection of the wide-band signal. From the above relations we have
R, œ Af", Ryo Af7 Af," so that

Ri Ais iac] ^
7 (ast) >n (2:58)

i.e., for a given receiver band, the range of detection of a narrow-band
signal is greater than the range of detection of a wide-band signal. 'The
two ranges A, and R, however, are not the maximum. In the former case,
the range can be increased by reducing the receiver band width to Afi,,;,
and in the latter it can be increased by broadening the receiver band width
to Afines We then have

Rem Rar) RR at) mem ban) 849)

Thus, despite the increase in the radiometric gain
with increasing band width, the maximum range of detec-
tion of narrow-band signals is greater than the maximum
range of detection of the wide-band signals. For exam-
ple, for Afe=10*Hz, Afin: =1 Hz, Afiwae= 100 Hz, we find R,=10°R,;

R max = 10* R= 316 Rz max.

EXTRATERRESTRIAL CIVILIZA TIONS

The dependence of the detection range on the band width in a system with
averagingis shownin Figure 43. Here, as before, we took 1, = rm i. e.,
I

the sender strives to attain the maximum transmission rate for a given
transmitter frequency band. The averaging action gives a gain in range if

M. The maximum range is attained for A;- Af,. These band widths
fall to the right of the line Af, = x, i.e., in the region where the radio-
2

metric gain is greater than unity only for t«v, when loss of information
occurs. For uv, the band widths corresponding to the maximum
detection range fall in the region without radiometric gain. Thus,
averaging produces a gain in range while resulting in a
complete loss of the information content.

I
TD
a zer
—— ———— ->
7
av AR nz SOC ic ors
Eu
s ar — e—— oe ee e a
oL
ES
4
J
2 Vide

4 -3 -2 -l 0 1 2 8 4 5 6 7 8 ysk

FIGURE 43. Range vs. receiver band width in a system with averaging. The
range Ris expressed in light years, P,g, = 200 dB: W, S:= 10* m’, T, "10K, B=!
T, =l sec. For t, < t,he range of detection at first increases with the decrease
of the receiver band width, as long as Af, > Af, and then, passing through

a maximum for Af, = Af, starts decreasing: this decrease stops for Af, = E .

when there is no radiometric gain (this is also the situation for Af; « Af,
without averaging). Fort, >t, the range of communication increases

~-
with a decrease in receiver band width in proportion to Afo E up to Af, = I
2

If this Af, is still greater than Af,, further decrease of the receiver band
width is accompanied by a more rapid growth of range (in proportion to

Afo 2), which stops for Af,= Af, The variation of range without averaging
is marked by the dashed lines in the figure.

This remark is applicable to the maximum range. It must be taken
into consideration in designing optimum communication channels, when
the receiver and transmitter frequencies may be taken equal. The
Situation is different in communication with extraterrestrial civilizations:
we cannot choose Af;—Af, since the transmitter band width is not known
in advance. Therefore, if the problem is not confined to the detection
of EC signals, but also includes the reception of the information contained
in the signals, the receiver band width should be chosen so that it is

II. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

not less than the expected transmitter band width, and the time constant
Should be taken sufficiently large to ensure an adequate radiometric gain.
However, tz must nevertheless be smaller than «,, i.e., we should have

aD ans Suppose there are grounds to believe that the transmitter band

width Af, is of the order of 0.1 Hz. Not to lose any information, we choose
the receiver band width with an adequate safety margin, Af,=1 Hz. Then,

to ensure an averaging gain, the time constant v; should be over 1 sec,

but not greater than 10 sec. Let us take 1,—4 sec, and the radiometric

gain will then be 2. We have thus obtained a slight gain, not in the detection
range, but actually in the communication range, i.e., the range of
information reception.*

The question of the assumed transmitter band width is highly uncertain.
We are never outside the domain of hypotheses on this topic. The band
width may be estimated from considerations regarding the most likely rate
of information transmission. If the rate of information transmission is
sufficiently low, the working frequency band is limited from below only
by the stability of the transmitted signal. In this case, the band may
reach a few Hz or fractions of Hz, and in molecular masers even
hundredths of Hz.

Range of reception of pulse signals

One of the ways for increasing the range of reception is through the use
of pulse signals. If the pulses are widely spaced, a sufficiently high pulse
power can be attained with a fairly low-power transmitter. Let Af; be the
pulse duration, and /, the time between two successive pulses. The ratio
of the instantaneous, or so-called peak, pulse power to the mean trans-
mitter power is a. To avoid averaging of the pulses by the recorder, the

1
time constant tz should not exceed the duration of the pulse. If this condition
is satisfied, the signal power P, is proportional to the power pulse. Each
sending of length A/, may constitute a simple or a complex pulse. In the
case of simple or so-called video pulses (without high-frequency filling),
the pulse duration Af, determines the transmitter band width Af, = a: In

3 1

this case, the condition t: < At, coincides with (3.45). Ift 24, the signal
power P,is proportional to the mean transmitter power. "Therefore, in
the range equations we should take

P, if Th,

P,=) Pit :
"Tp AM f egat (3.50)

If n». e., when the sender radically reduces the quantity of information sent through the channel
1

in unit time), the receiver band width Af, and the time constant t2 can be chosen so that reception of

information is ensured for a sufficiently large radiometer gain of the order of V v, Af,, i.e., in this

case, no information is lost in averaging.

EXTRATERRESTRIAL CIVILIZA TIONS

The maximum range of communication is attained for Af,—-Af,, and since in
this case, as we have seen, wAf,=1 (for s <44), we obtain

= Pigits, A
Rmax = ( dna MEARE) . (3.51)

In particular, for simple pulses, when AAAf,-1,

1,
Rmax = (AES), (3.52)
i.e., the maximum range of communication using simple pulses and
a fixed mean transmitter power is independent of the transmitter band
width and increases with the increase in the time spacing between the
pulses. The feasibility of high-range communication with a relatively
low-power transmitter* using widely spaced pulse signals makes this
communication technique particularly attractive for interstellar com-
munication. Although the wide pulse spacing lowers the transmission
rate of the system, the loss of information is not particularly significant
for the transmission of call signals by extraterrestrial civilizations.

Length of transmission. Directivity and information
content

As the directivity of the receiving and the transmitting antennas is
improved, the signal-to-noise ratio and the resulting range of communication
both increase, Should we thus always strive to increase the directivity of the
transmitting antennas to the maximum?

Let us consider the relationship between the length of transmission and
the directional properties of the antenna, when the exact position of the
subscriber is not known in advance. This situation is a good approximation
of what we are likely to encounter in communication with extraterrestrial
civilizations. Suppose the transmitting antenna is mounted on a planet
which, like the Earth, spins with a rotation period Tp and let the antenna
axis remain fixed in the planetary system of axes. As a result of rotation,
the directional radio beam radiated by the antenna intercepts any given part
of the sky for a limited length of time A/, as long as the corresponding
sky area falls inside the main lobe of the antenna pattern. The faster
the planetary spin and the higher the antenna directivity, the shorter is
the time A/. Suppose that at some time t the antenna is aimed exactly

at the subscriber. In a time 4, it rotates through the angle

At
> (3.53)

where Bao, cosó is the velocity of rotation of the beam, wp is the angular

velocity of rotation of the planet, 6 is the angle between the antenna axis
and the plane of the equator. Signals are received at the relevant time

* For numerical estimates see Tables 3.3— 3.5 and Figure 45.

100

Hl. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

RNC Lx if the angle of rotation z ® does not exceed half the beam width. To

fix our ideas, we may set the maximum angle of rotation for which the
signals are just received equal to the beam half-width between points of

half power (s = $5). Then the total transmission time At (from tt to
t+ F) is

At = es, (3.54)

Op cos 6

Consider a pencil-beam reflector antenna. The dependence of the
beam width on the directivity coefficient for this antenna is expressed by
the relation

62g = const, (3.55)

The numerical value of the constant depends on the exact power level
used in reckoning the angle 0. If 0=20).5, then const = 10.2; if 0—26,,
i.e., the total width of the main lobe (or the beam width between zero
power points), we have const = 59.2. Using this dependence, we can
establish a relationship between the length of transmission Af and the
directivity coefficient of the antenna. Figure 44, borrowed from
Webb /7/, gives some idea of the value of A! for various g in the case
of a planet with a 24 hr period of axial rotation, when the antenna axis
is aligned in the equatorial plane.

Antenna gain

inst

a LI) TUL LL TII
o X ou m 10
Time, sec

FIGURE 44. Directivity of transmitting antenna as a
function of the length of transmission.

The antenna is fixed in the planetary system of axes.
The antenna axis is aligned in the equatorial plane,
and the planetary rotation period is 24 hrs.

Suppose that after each complete rotation of the planet the antenna axis
is displaced in declination through an angle 6 equal to the beam width, so

101

EXTRATERRESTRIAL CIVILIZATIONS

that ever new sky areas are illuminated after each rotation. The total
time to scan the entire sky is

AT-T,S ~ T, VE, (3.56)

and the length of transmission received by each subscriber is expressed
in terms of the total scanning time in the form

AF
At = -si (3.57)

The minimum duration At = apis observed for subscribers located in the

equatorial plane. A similar relationship between the total scanning time
and the length of transmission to each subscriber is obtained for cases
when the scanning is done by moving the antenna proper, without resorting
to the planetary rotation; in this sytem, the antenna tracks for a time At
one given sky area, and is then abruptly aimed at the next area.

Let us now consider the relationship between directivity and information
content. Let the transmitter power P, and the length of the transmission
AT be given; let P, be the power signal at the reception point at a given
distance R in the case of isotropic transmission. The length of trans-
mission for each subscriber in isotropic sending is equal to the total
length of transmission. Therefore, the maximum quantity of information
Q, that can be transmitted in this time by an isotropic transmitter is given
by Shannon's theorem:

Qı = A AT log; (1 + $2). (3.58)

Let us now consider directional transmission with the same P, and AT.
The signal power is increased by a factor g due to directivity, and the
time of transmission toward each subscriber decreases by the same
factor. 'Therefore, the maximum quantity of information that can be
transmitted in a time AT by a directional antenna, when the subscriber's
position is not known in advance, is given by

Q, = Af at log, (1 + 5+) = af ST tog, (1 + £75). (3.59)

Comparison with (3.58) shows that

Qi Qi. (3.60)

The equality is observed when
Ps gP
Bo <p, SL (3.61)

In this case, changing over from binary to natural logarithms and series
expanding, we find

LEPAT O P.AT
Q= ndgPn T mPay 7 O (3.62)

102

III, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

i.e., for $ « 1, the quantity of information grows in proportion to the
n

length of transmission and the ratio of the signal power to the specific
noise power; it is independent of the channel band width.

Equation (3.60) shows that if the required direction
of transmission is not known beforehand, the quantity
of information decreases on passing from isotropic
transmission to directional transmission, all other
conditions remaining constant. This conclusion was derived
by Siforov /8/. 'The decrease of information content is associated with
the uncertainty in direction to the subscriber. If this uncertainty is
reduced, the situation changes radically. Suppose that various considera-
tions (astronomical or other data) indicate that the sending EC should be

sought in certain directions in space only. Let + be the ratio of the total

solid angle Q corresponding to these "civilized" directions to the entire
solid angle 4m. Directional antennas will be more advantageous if

X gPs P
y log, (1 + $+) > logs (1 + $2). (3.63)
In particular, for P =P, and g>1, we have

Q
aoa! QG&y <4 lg g. (3.64)

For g= 10°, the size of the region to be scanned with directional trans-
mitting antennas should not exceed the beam area (between points of half
power) by more than a factor of 24. For weak signals, the uncertainty in

direction may be increased. Leta= 5 « but agz»1. Then,
n

0.43ag Q 3Iga,
Y? gap? Bah Sa 6569]

For g=10°8 and «—10 7, the region of uncertainty (the search region)

may reach 1000 times the beam area.

$3. CALL SIGNALS AND ARTIFICIALITY CRITERIA

Before establishing communication with extraterrestrial civilizations,
we should first detect the sources of artificial signals in outer space.
The main difficulty is not that these signals must be picked up against the
background of cosmic radio noise (a similar situation is observed in
ordinary radio and radar systems): it is that the sources of these
signals must be reliably identified and distinguished from a tremendous
number of natural radio sources, such as galaxies, radio galaxies,
quasars, ionized and neutral hydrogen clouds, supernova remnants, and
even individual stars.

To isolate the meaningful radio signals from the jumble of radiation at
the receiver input, the incoming radiation must be appropriately

103

EXTRATERRESTRIAL CIVILIZA TIONS

processed. This processing depends on the method of modulation employed
by our counterparts. Modern technology provides us with a wealth of means
for the analysis of radio waves, but there is no point in applying these
analytical tools to sources whose natural origin does not raise any doubts.
Before proceeding with the actual analysis, we have to establish that we
are dealing with an artificial radio source, or at least there is enough
evidence to suspect a source of artificial origin. It would therefore seem
that the radiation from an artificial source would possess some peculiar
features intended to simplify its detection and identification by other sub-
Scribers. Hence the need for a sort of call signal from extraterrestrial
civilizations.

We can advance a number of assumptions regarding the likely compo-
sition of EC call signals. First, they should ensure a high detection re-
liability. This condition is best achieved with the aid of continuous
(although possibly variable) radio transmission. If the subscriber's
position is not known in advance, the transmission should be isotropic,
since a highly directional transmitting antenna scanning the sky produces
a very short transmission in every given direction (see Figure 44). It
moreover seems likely that the call signals contain some information
regarding the artificial character of the source, indications of frequency
and band width of the transmission, and some additional information
which may be regarded as a "key" to the main program. The overall
quantity of this information is not particularly large. Therefore, narrow-
band quasimonochromatic signals will do as call signals. This is
a highly advantageous turn of events, since, on the one hand, a long
range of communication is ensured and, on the other, the artificial
Source can be identified with fair certainty. Indeed, the great majority
of the natural radio sources show a very wide, almost unbounded, con-
tinuous spectrum. Even the monochromatic radiation of interstellar
hydrogen at 21 cm fills a fairly wide band of the order of 5-104Hz. The
narrower band of the 18cm hydroxyl emission, which is assigned to a
natural maser mechanism /9/, isa few hundreds of Hz wide. These narrowest
natural band widths are clearly inferior to artificial signal generators, which
provide band widths of a few Hz or even fractions of a Hz; molecular
masers emit in band widths of a few hundredths of Hz. The very
detection of such narrow-band signals in itself would provide an indication
of a possible artificial origin of the source. Note, however, that the use
of narrow-band signals leads to certain difficulties associated with
frequency scanning. This problem, however, is not insurmountable,
and it will be discussed in the next section.

Along with the narrow-band quasimonochromatic signals, we can
expect call signals in the form of widely spaced pulses. This approach
also ensures a long range of communication and clearly labels the signal
as artificial: natural radio sources generally emit continuously.* Special
equipment is required for the detection of these signals.

Although the application of special (narrow-band, pulse, etc.)
signals as EC call signals seems to provide the most logical and likely
approach to the problem, we cannot rule out another possibility, namely
that the transmission will be continuous in a wide frequency band (to
ensure a high rate of information transmission), and the function of call

* A remarkable exception to this rule are the pulsars.

104

Ili, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

signals will be fulfilled by the properties of the source itself and special
features of the continuous transmission.

We thus have to solve the problem of the criteria of artificial origin
of radio sources. This topic was first attacked by Kardashev /10/.
Later it was analyzed by Slysh /11/, Gudzenko and Panovkin /12/, and
others. The proposed criteria can be divided into two groups:

1) criteria or signs following from the artificial origin of the source;

2) special properties of radiation, intentionally imposed by the sending
EC to ensure communication and simplify detection.

The first group includes such features as angular dimensions,
spectrum, statistical properties of signal, variation associated with
possible rotation of the system. The second group includes circular
polarization, variation associated with modulation, information regarding
artificial origin and "keys."

The angular dimensions are one of the most promising and indicative
criteria of the first group. The angular size of artificial radio sources
cannot exceed a certain (fairly small) value. On the one hand, this is
related to the limited scale of activity of civilizations in space (e. g.,
the scale of a planetary system) and, on the other hand, to the finite
Speed of propagation of information. Indeed, let ! be the time between
two successive pulses. 'To ensure simultaneous emission from different
parts of the transmitting system, the distance between the different
parts and hence the linear size L of the entire system should not be
greater than ct, where c is the velocity of light. If R is the distance to
the transmitting system, its apparent angular dimension is

e c
9X TW. (3.66)

where q is the rate of information transmission. For a distance of 1 kpc
and g= 3-107* (which corresponds to a transmission of one bit of information
per hour), we find 9 < 0".007. As the rate of information transmission
increases, the maximum angular dimension of the source correspondingly
decreases. For a rate of 1 bit/sec, q« 0.000002. The angular dimensions
of natural sources are generally much larger. Even the less extended
sources (the source of the OH line) have angular dimensions of the order of a
few thousandths of an angular second. When the steady increase in the
sensitivity and the resolving power of radio telescopes will enable us to
pick up radio waves from individual stars, this criterion will of course
lose some of its paramount importance, but in combination with other
signals (power, band width, etc.) it will probably retain much of its value.
If the EC transmitter sends in a sufficiently wide frequency band, its
radiation will not be unlike the continuous emission of an artificial source.
However, the spectral power distribution of the transmitter will probably
differ from the power distribution in the spectrum of natural radio sources.
This topic was treated in detail in Chapter I. If the aim is to ensure a
maximum transmission rate, the spectrum of the artificial source should
look like the curves in Figures 21 and 22. A curve of this shape may be
accepted as one of the criteria of artificial origin. This criterion, how-
ever, is not very decisive. First, the condition of maximum trans-
mission rate is not absolutely binding. Moreover, excessive saturation
of the signal with meaningful information is undesirable, as it interferes
with decoding. Second, a similar power spectrum curve may be observed

105

EXTRATERRESTRIAL CIVILIZATIONS

in some cases for natural sources also. All this notwithstanding, this
criterion has its value. In combination with other properties of the radio
waves, it may prove to be very useful in establishing the exact nature

of the source.

The same considerations apply to signal variation associated with
possible rotation of the system. In this case, the length of transmission
is determined by the period of rotation and the directivity of the trans-
mitting antenna; the total period of power variation, however, is entirely
determined by the rotation period. Variations with periods from a few
hours to several days can be expected for transmitters mounted on a
Spinning planet, and variations with periods from a few months to a few
years should be observed for planets or other celestrial bodies which do
not spin and only travel around their primary, at a certain distance from
it (in the corresponding "zone of life"). For a long time, the opinion
prevailed that natural radio sources have a high degree of power constancy.
This conclusion emerged from theoretical calculations and there was
ample observational evidence to support it. However, after the discovery
of the variable radio source CTA-102 /13/, the situation changed radically,
since this discovery was soon followed by the detection of the variable
radio emission of quasars at various frequencies and with various
characteristic times (from a few days to several years). This criterion also
has lost its paramount importance, but like the other criteria it should be
kept in mind.

The strongest criterion of the first group is apparently that associated
with the statistical properties of radiation. This topic was considered by
Golei /14/, Slysh /11/, Gudzenko and Panovkin /12/, and Siforov /8/. The
radio emission of natural sources is a random, uncorrelated noise, since
it is made up of a multitude of independent elementary emission events.

In artificial signal generators, on the other hand, the individual emission
events are not entirely independent. Therefore, the statistical properties
of artificial radiation (e.g., the amplitude distribution) are different from
those of noise. The search for artificial radio sources
should therefore provide for a comprehensive analysis
of the statistical properties of signals.* Analysis of this
kind for very weak radio sources is a formidable undertaking. It requires
Special equipment, different from the conventional tools of the radio
astronomer. Note that although the need for a greater emphasis on the
statistical analysis of signals has been stressed, little has been done in
this direction.

Let.us now consider the criteria of the second group. Plane-polarized
radiation propagating in the interstellar medium may experience a pro-
nounced rotation of its plane of polarization in the interstellar magnetic
fields as a result of the Faraday effect. This is a common phenomenon
in radio astronomy, and it is often applied to estimate the distance of the
radio source from the observed rotation of the plane of polarization.
Although in radio astronomy this is a useful effect, providing additional
information regarding the radio source, it is highly harmful in connection
with the problem of EC communication, as it definitely distorts the
incoming information. The Faraday effect is a sensitive function of frequency.

* The statistical analysis can be based on the moments of the distribution function, the autocorrelation
function, the spectral correlation function. etc. /11/.

106

IIl. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

Therefore, different spectral components of a wide-band signal undergo

a different rotation in the interstellar magnetic fields. As a result, the
antenna, responding to one direction of polarization only, will record

the different spectral components of the signal with different attenuations. The
Spectrum will be distorted, and the true time characteristics of the signal
will become unrecoverable. To avoid this unpleasant effect, the radio
transmission sent by the EC should be circularly polarized to start

with, and it should be received by a circularly polarized antenna. This

is indeed the practice in long-range space communication systems

in the solar system.

Variation due to modulation is the most reliable sign of artificial
origin of radio signals. The main difficulty, however, is that the
characteristic time of the probable power variation is unknown. If the
modulation is associated directly with information encoding, the modulation
time is probably very short. In binary transmission, with transmission
rates of 1000 bit/sec (this is hardly a high rate of transmission: television
requires a thousand times higher rate), the characteristic time of power
variation, which coincides in this case with the duration of the binary
pulse, is 107? sec. To record such fast variations, we need special
equipment with a very small time constant 1 «10 -3 sec. To ensure high
Sensitivity despite the small time constant, we have to use antennas
with a very large effective surface.

The rate of information transmission of call signals may be much
lower. Rates of 1 bit/sec are probably more than enough in order to
transmit the few tens or hundreds of bits of information probably
contained in call signals within a reasonable time. As regards the exact
nature of this information, intended to announce the artificial origin of the
radio source, we can only guess. Some suggest that several natural or
primary numbers can be transmitted to this end; others prefer combi-
nations of known mathematical constants, such as e and m.

Note that special monochromatic signals are not the only candidate for
call signals: wide-band signals generated by modulation of short informa-
tion-carrying pulses will also do. The signal variation, in this case, may
correspond to interruptions in the main program, e.g., the beginning or
the end of a certain transmission session. These slow power variations
of wide-band radio signals can be detected with the existing radio-
astronomical equipment. However, it is very important to know the
expected period of variation: is it seconds, minutes, or years? This
question cannot be answered at this stage. We can only fix a rough
lower limit for the probable characteristic time of power variation in
the EC call signals. If the transmission is conducted at a frequency v,
the modulation time q in the EC call signals should satisfy the inequalities

t>q =A >v], (3.67)

For radio frequencies, this gives t>107!! — 107? sec. A more exact
estimate can be obtained from the requirement of pulse stability during
propagation in the interstellar medium. We have seen in Chapter II that
the group delay effect associated with differences in the group velocity
for various quasimonochromatic wave groups making up the wide-band
pulse imposes certain restrictions on the pulse duration t. Thus, for a
galactic source operating in the range of decimeter wavelength, the pulse

107

EXTRATERRESTRIAL CIVILIZATIONS

duration should be much greater than 10 ^8 sec (if the source lies outside the
plane of the Galaxy) and much greater than 107? sec (if the source lies in the
galactic plane). For an extragalactic source, the limiting pulse duration
may reach 107* sec. Anyhow, we may write

1710? sec. (3.68)

The problem of detecting EC call signals would be essentially simplified
if we could fix a standard modulation period likely to be used by all EC.
This period should naturally satisfy conditions (3.67) and (3.68). We can
try to approach this problem by choosing an appropriate combination of
universal constants which has the dimension of time or taking as our basis
the characteristic time of some processes which are common for the
entire Universe, e.g., atomic or comological processes. One of such
possibilities is the atomic unit of time equal to the period of orbital
revolution of the electron in Born's first orbit, or the so-called Jordan
elementary time, equal to the classical radius of the electron divided
by the velocity of light. The former quantity is equal to 2.4* 10" sec,
and the latter to 9.4*10 7 sec. However, none of these times satisfies
(3.67) nor (3.68). These units of time fix the time scale of microcosmic
phenomena. They can be called microscopic time units. On the other
hand, there is a completely different megascopic time scale associated
with the expansion of the Universe, the time scale characterized by
Hubble's constant H, the universal megascopic constant. It would seem
that the modulation period in EC call signals should logically fall "half-
way" between the microscopic and the megascopic time units; for example,
it may be chosen as the geometrical mean of the corresponding numerical
values. 'Taking the same atomic and Jordan elementary time and using
the megascopic unit H-!= 3.101" sec, we obtain two macroscopic time units

2,—V2.4 107 x3. 10" =3 sec
and

t.=V9.4-10%x 3-10" = 0.002 sec.

Both these values satisfy conditions (3.67) and (3.68). A logical micro-
Scopictimeunitis v’, where v is the frequency of the transmitted signal.
For the optimum frequency range of interstellar communication

v'=1011 — 107? sec, and the corresponding macroscopic times fall between
30 min and 4.5 hours. The above examples are clearly very sketchy.

In particular, the difficulties associated with exact determination of
Hubble's constant make it highly unsuitable for use as a basic time unit.

It would appear, however, that extraterrestrial civilizations contemplating
interstellar communication should have a sufficiently accurate knowledge
of it.

We should probably start looking for variations with a period of a few
hours. These variations are fairly easy to detect, since no special equip-
ment is needed. These long-term power variations are not distorted by
shimmering effects which accompany the propagation of electromagnetic
waves in the interstellar and interplanetary medium and inthe ionosphere.

108

IIl. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

The time scale associated with these variations is fairly characteristic of
the macrocosmos to which our partners apparently belong (at least if we

are dealing with anthropomorphic civilizations). Finally, the very discovery
of periodic power variations of period t related to the radiation frequency

by the equality t= V 4 !H^! would attract enormous attention to the correspond-
ing effect.

In conclusion of this section, we would like to stress that the entire
topic of call signals and artificiality criteria has hardly been studied so far.
Much that is unclear and uncertain remains in this field, opening wide horizons
for future research. Rigorous and single-valued criteria should be developea
for identifying artificial sources. Such criteria can be based, e.g., on the
analysis of the statistical properties of signals or on general theorems
of information theory and the theory of complex systems. Some guidelines
toward the solution of this problem are indicated in Chapter VI.

$4. METHODS OF DETECTION OF EC SIGNALS

Transmitter power. The power potential of a civilization

In our search for EC signals, we are faced with a two-fold uncertainty:
we do not know at what frequency and in what direction these signals are to
be sought. A similar uncertainty is in force for the sending EC. The
simplest solution to this difficulty is to set up continuous transmission
of sufficiently wide-band signals in all directions in space. This ensures
simultaneous "service" to all the civilizations within the sphere of action
of the transmitter and enables new subscribers to tune in as soon as they
reach a suitable technological level. If the signals are sufficiently powerful,
and the receiver has a sufficiently high sensitivity, the signals can be
received with low-directivity or even isotropic antennas. This has con-
siderable advantages, as it eliminates the need of direction scanning in the
first stages of detection. However, this "simple" communication system
requires tremendous power. Table 3.2 lists the minimum transmitter
power needed for detection and communication using continuous isotropic
transmission and undirectional reception at 3 cm wavelength. For detection
purposes, the effective signal is assumed to exceed by a factor of 10 the
rms noise fluctuation (p=10), and for the purposes of information reception,
the signal is supposed to exceed the noise level by a factor of 100 («—100);
the noise temperature was taken equal to 10°K, and the time constant (for
detection) *;— 100 sec. Finally, in accordance with the conditions of
minimum power, we took Afi=Af2=Af, The band width Afis expressed in Hz,
the distance R in light years, the transmitter power P, in watts. As we see
from the figures in Table 3.2, the required power not only falls far beyond
the possibilities of the current transmitters, but actually exceeds the
total power potential of mankind.

Mankind is currently consuming annually about 1.5 * 10?" erg of energy
of various forms, which corresponds to a power of about 5 * 10!? watt.

109

EXTRATERRESTRIAL CIVILIZATIONS

TABLE 3.2. Minimum transmitter power for continuous isotropic transmission and nondirectional reception at
3 cm wavelength (g;— g27 1; Af; = Ma = AP)

Detection Communication (reception of information)
(B= 10; t,= 100 sec; T, = 10*K) (o = 100; Ta = 10°K)
^
d 10 (o n 1 10 | 1 10 10? 10 10 10 os m
R
10 |2.10'7|2- 1018 2. 10!? 2- 102 2. 10?! 2- 1074/2 - 1019 2- 107° 2- 102) 2- 10?? 2. 1023 2. 1074 2. 1025 2- 1076

10 [2-109 2. 10% 2-10? 2- 1022 2.1023 2- 10%]2. 1021 2- 102? 2. 10% 2.10% 2. 10% 2. 10**[2 - 107 2-10
- 10? 2- 102 2. 102? 2. 10% 2. 102 2- 10792 - 10232. 10% 2. 1075 2. 10262. 10? 2. 10?* 2- 1029 2. 1990
10% 2- 10 2. 1022 10**[2 107 2 - 10?*- 10% 2- 10°62. 107 2.10% 2. 1029 2. 10% 2. 10% 2. 10%
- 10552. 1026[2- 107 2-10 2- 102 2- 10°12. 1027 2- 102 2. 102 2- 10° 2. 10%! 2. 109? 2-108 2. 10%

- 1027 2- 10% 2. 1079 2- 10% 9. 10% 2. 1032. 1029 2- 102° 2- 10% 2- 109? 2. 1053 2 . 10% 2- 1055 2 . 10%
10 |2- 10292. 109? 2. 10?! 2- 102 2. 1033 2- 10*2 . 103! 2- 1082 2- 10% 2-10% 2. 10% 2. 1035 2- 10772 - 1095
10* |2- 10°! 2- 10% 2.1092. 10% 2.103 2. 102. 105 2- 10% 2 - 1055 2-10% 2. 1057 2-10 2-10 2- 10%
10 |2- 108 2- 10% 2. 10% 2- 10% 2 10°] 2- 10%|2- 10552 - 10% 2. 10*7]2- 10% 2- 10% 2.10% 2. 109 2. 108

-
e
a
NIN N N

The entire energy consumed is eventually degraded to heat and then radiated
into outer space. In principle, it could be convered into radio waves
(this does not clash with the thermodynamic laws) and then used for inter-
stellar radio communication. However, the entire power would not be
enough for ensuring continuous isotropic transmission aimed at an un-
directional receiving antennas within the range of a few tens of light years,
i.e., the message would not reach the nearest stars. This does not mean,
however, that this convenient method of communication is completely
hopeless. Since we assume that our civilization is not unique in the
Universe, it inevitably follows that there should be civilizations on a lower
technological level than ours, on the same level with us, and of course on
higher levels of development. The highly advanced civilizations may have
tremendous power resources at their disposal, which are absolutely in-
accessible to mankind at the present stage of development. The power
potential of a civilization in the last analysis determines the power of its
transmitters. On the other hand, this is one of the most important
parameters of interstellar communication affecting the range of detection
and communication, the quantity of transmitted information, the kind of
signals, used, and, indirectly, the methods of detection. Therefore, the
question of the probable power potential of a civilization merits a more
detailed examination.

The main features of the growth of the principal indices of technological
progress of civilizations are analyzed in Chapters I, V, and VI. We will
consider here only the growth of the power resources of a civilization.

The annual growth of power consumption in the world is about 3% during
the last 100 years. Ifthe same rate of growth will persist in the future,
the per-second power consumption on the Earth will reach 10!" watt in the
next 300 years, thus becoming equal to the influx of solar energy. Further
increase of power consumption will be unfeasible, since this will radically

110

IH. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

change the radiation balance of the planet.* This is a highly significant
factor, whose importance is generally underestimated. It should be
emphasized that this restriction of power output has nothing to do with
shortage of power resources: it follows from the necessity to maintain
the equilibrium in the atmosphere and on the surface of the Earth.

A similar situation is encountered by any civilization on any planet.
Since the energy received by a planet in the "life zone" from its primary
cannot change between very wide limits, the energy output of any planetary
civilization should be limited by figures of the order of 10!" watt. When
this limit is reached, further development is possible only through active
conquest and population of interplanetary space, where high-power
installations and industries should be moved. In the light of this con-
clusion, it seems that the exploration of space which has recently begun
is a vital step toward ensuring the future growth and existence of our
civilization, and by no means can it be regarded as premature. Active
conquest and population of outer space will eventually lead to the
creation of an artificial biosphere around the Sun (the Dyson — Tsiolkovskii
sphere). A civilization of this kind, inhabiting an artificial biosphere around
its primary, should have access to much higher power outputs, reaching
3:10% watt. Assuming an exponential growth, the time to reach a Dyson-
type civilization is not very long. Indeed, if the annual growth of power
output is merely 1%, the transition from a civilization with an energy
output of 10" watt toa Dyson civilization with energy requirements of the
order of 3:107? watt should take about 2200 years. After another 2500 years,
the per-second power consumption, assuming the same growth rate, will
reach 10?" watt, which is equal to the radiation energy of all the stars in the
Galaxy.**' This extrapolation of the growth of civilizations can be extended
into the more distant future, but we had better stop here. In 1964,
Kardashev /10/ proposed a division of allthe civilizations into three types
in terms of the power requirements. Type I civilizations are those which
are close in their technical development to the Farth civilization (power
requirements of 107? — 107? watt), type II civilizations are those with power
requirements of the order of 3- 1076 watt, and finally, type III civilizations
are those which have harnessed the power resources on a galactic scale,
with energy output of 1037 watt, We will follow Kardashev's terminology,
but extend the concept of type I civilizations to all planetary civilizations
with power requirements close to the Earth level and higher, up to 10! watt.
The existence of supercivilizations with energy requirements of the order
of 1026 — 1037 watt is a mere hypothesis. However, strictly speaking, the
very aSsumption regarding the existence of other extraterrestrial civili-
zations is also a mere hypothesis. It is therefore advisable not to ignore
any of the possibilities.

Let us return to Table 3.2. For a band width of over 1 Hz, detection
of signals and reception of information from type I civilizations is ruled
out even for the nearest stars. Thus, only type II or type III civilizations
can communicate by means of continuous isotropic transmission with un-
directional reception. With bands of 1 MHz, the detection of signals from

* Actually, the power output level will have to be frozen at a much earlier stage, when it reaches a few per
cent of the solar energy flux received each second (i.e., in about 100—200 years).
** For a higher annual growth rate, these limit values will be attained much sooner (see, e.g., Chapter I,
p. 26),

EXTRATERRESTRIAL CIVILIZATIONS

type II civilizations is possible over distances of 1000 light years, but
reception of information is possible only over distances corresponding to the
nearest stars. For band widths Aj «100 Hz, signals from type II civilizations
can be detected anywhere in the Galaxy, whereas reception of information

is possible over distances greater than 1000 light years. The signals of

type III civilizations can be detected virtually anywhere in the observable
Universe. For band widths of 10 MHz, information can be transmitted only
to the nearest galaxies, whereas for sufficiently narrow bands, Af« 100 Hz,
information can be transmitted within the limits of the Metagalaxy.

Hence it follows that if at least one type II civiliza-
tion exists anywhere in our Galaxy or at least one type
III civilization exists anywhere in the Universe, and
these civilizations using their tremendous power
potential transmit continuously in all directions power-
ful monochromatic signals of band width Af<100 Hz, we
Should be able to detect these signals even without
knowing where the source is located.

This method of communication is the least advantageous in terms of
power. Other methods of transmission and reception require much lower
power levels. Consider a high-quality receiver with noise temperature
of 10°K and band width of 100 MHz which functions at 3 cm wavelength.
Suppose that this receiver is placed outside the atmosphere, where the
total noise temperature is determined by the receiver noises, being equal
to T, 210?K. Let us now determine what power is needed for the detection
of signals or reception of information over distances of 1000 light years
for various reception and transmission techniques. We have chosen the
distance of 1000 light years because, according to some modern estimates
/2,3,15/, this is the average distance to the nearest EC. "Table 3.3 lists
the power values (in watt) necessary for signal detection and communication
over distances of 1000 light years assuming fı < Afs. An area of 10* m?
was assumed for the receiving antenna, which corresponds to the area of
the largest modern radio telescopes in the centimeter and decimeter range.
For directional transmission, the table gives the product of power times the
antenna gain in dB- W and also the power corresponding to the gain g;— 10?.
The dashes in the last two columns indicate that, formally, the condition
of pulse signal detection (tz < Aj) coincides with the condition of information
reception (see $2). Note that for the problem of communication with EC,

the main factor is not the relative pulse duration m, but the length of time
1

t; between the successive pulses, which determines the rate of information
transmission by pulse signals. Since in our case At, = A» 10-9, the
relative pulse durations listed in the table correspond to the following in-
formation transmission rates:*

Af = 197, 107, 107, 107, 1075, 107, 107”
1
t2 107, 1075, 1075, 107, 107°, 1, 10° sec

e-1- «10, 105, 105, 10%, 107, 1, 107? bitysec.
1

* On the assumption that binary pulses are used. If a pulsed code with some base a2 is used, the
transmission rate figures should be multiplied by log; a.

112

Il. RADIO COMMUNICATION WITH EXTRA FERRESTRIAL CIVILIZATIONS

TABLE 3.3. Transmitter power needed for the detection of narrow band signals and communication over
distances of 1000 light years (A= 3 cm, Af, < Afa = 100 MHz, T, = 10*K)

Communication; e - 100 Detection; 8-410, t:= 100 sec
Transmission nondirectional |directional reception] nondirectional directional reception
reception S,- 10* m*; reception S,» 10* m?;
T, = 10°K T, = 10°K T, = 109K T, = 10°K
Isotropic continuous 2:103! 1.5710? 2:105 1.5: 107
Isotropic pulse
bt
ti
107! 2:109 1.510?
107? 2-107 1.5. 10?!
1074 2-10°7 1.5- 101? =
10-5 2:195 1.8: 10!7
1078 2:10? 1.5: 1015
10-9 2.10! 1.510!
Directional continu- | | Pg, P, for Pigs P, for P, for
ous dB.W ge 109 | dB.W g,- 10° dB.W — g-10?
313 2.107 | 232 1.5.10% 1.5.10?

Directional pulse

EOM
ly
107 303 2:10?! 222 1.5:10!
107? 203 2.10? 212 1.5-10”
10-4 273 2-10? 192 1.5:10??
1075 253 2-10" 172 1.5.10* -
10^? 253 2-10% 182 1.5.109
10720 213 2-104 132 1.5-10*

We see from Table 3.3 that type I civilizations will remain undetected
at a distance of 1000 light years in the case of continuous isotropic trans-
mission and nondirectional reception. Signals from type II civilizations
can be detected under these conditions, but there is not enough power for
the reception of information. To ensure information reception, we should
switch over either to pulse or to directional transmission or, alternatively,
to directional reception. In the case of continuous isotropic reception and
directional reception, we can detect signals from type I civilizations
and receive information from type II civilizations over these distances.

t E
Isotropic pulse transmission with relative pulse duration (F4)< 1075

makes it possible to establish communication with type II civilizations using
a nondirectional receiving antenna. The transition to a directional re-
ceiving antenna with an effective area of 10* m? makes it possible to establish
communication with type II civilizations for almost any relative pulse
duration. For (^2) «1075, information can be received from type I civili-
zations (at transmission rates lower than 100 bit/sec); in particular, for

At ; A E z "
(45) =10-%, when the rate of information transmission is higher than

1 bit/min, power of the order of 10 watt is required, and this figure is

113

EXTRATERRESTRIAL CIVILIZA TIONS

comparable with the present-day power output of mankind. In the case of
continuous directional transmission and nondirectional reception, we can
detect signals from type I civilizations and receive information from type II
civilizations. In case of directional reception and transmission, 10!! kW
are required for the reception of information and 150 MW for the detection
of signals. Directional pulse transmission makes it possible to establish
communication with a type II civilization using a nondirectional receiving
antenna for pulses of any relative duration. For relative pulse duration
of less than 1075, communication with type I civilizations is possible if a
nondirectional receiving antenna is used; with a directional antenna having
an effective area of 10* m?, communication with type I civilizations is
possible for any pulse duration (continuous transmission included). Finally,
in case of continuous pulse transmission with relative pulse duration of less
than 10-8 and reception with a directional antenna of effective area of 10* m?,
a mere 1.5 MW is needed

Table 3.4 lists the power values required for the detection of signals with
a communication system with the same parameters assuming Af, Af».
Note that in this case we can only discuss signal detection, since for Afi Af;
communication inevitably involves signal distortion and loss of information.
This should be kept in mind in reference to the left half of the table, which
lists the power values required for this incomplete communication. In
distinction from the previous case (Afi<Afe), the power in pulse transmission
now depends on the time spacing between the pulses, and not on the relative
pulse duration. When comparing the data of Table 3.4 with the previous
data of Table 3.3, we should remember that since now At, = Af; < 107° the values
of fj listed in Table 3.4 correspond to the following relative pulse durations:

f, I5 jm j^ 24^ | month !year

A et 2.107? 3.1077 107? 4.1075 3.107%
1

Table 3.5 lists the minimum power values required for communication
over distances of 1000 light years assuming equal receiver and transmitter
band widths. Examining this table, we readily see that in case of directional
reception and transmission, detection of signals and communication can be
established over distances of 1000 light years with very small power,
especially if pulse transmission is used. From the point of view of power
requirements, this is the best method of communication. However, the
detection of these signals is very unlikely, unless the direction of transmis-
Sion and reception are known in advance. Isotropic transmission should be
used, as we have noted before, to enable new subscribers to tune in. If
the transmission is continuous in time, it is better to use directional
receiving antennas subsequently scanning different parts of the sky. This
procedure requires power of the order of 10% — 107" watt, which is
available to type I civilizations. In case of pulse signals, especially
when the pulses follow one another at large intervals, it is better to use
a nondirectional antenna, since in this way a continuous sky survey can be
conducted and the probability of detection markedly increases. This
approach, however, requires power of the order of 107? — 10? watt. The
lower of these figures corresponds to very slow transmission (1 bit per
year), so that we should actually speak of signal detection, and not
information reception. (There is a possibility, however, that transmission

114

Ill, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

rates of the order of 1 bit per hour or even 1 bit per year are acceptable
for call signals.) An even better way to search for pulse signals is with
the aid of a system of directional antennas, which jointly cover the entire

sky.

suffice for the detection of pulse signals.

If each antenna has an area S;— 10* m?, power of 107 — 10% watt will

TABLE 3.4. Transmitter power needed for the detection of wide-band signals in communication (with partial
loss of information) over distances of 1000 light years (& 23cm, Af, > Af: = 10° Hz, T, = 10°K)

No.

Transmission

Isotropic
continuous
Af, = 10?
Af, = 1010

Isotropic
pulse f, =
psec
1 min
hr
4 hr
1 month
! year

Directional
continuous
Af, = 109
Af, = 1019

Directional

ulse f, =
p pec!

min
y hr

24hr
| month

__l year,

Communication (with loss of information)
a~100

nondirectional re- |directional reception, | nondirectional re-
ception, T, = 10°K |s,210'm?; r, =10°K| ception, r, —10*K

2.102 1.5- 102 2.1025
2- 1033 1.5. 1025 2.1077
2-107 1.5. 1015
3- 10?! 2.5. 105
6 - 101? 4-10! zs
2- 10!8 2- 10!
8- 1015 6- 108
7.1015 5.107
Pigi P, Pun P, Pigs Py
dB: W for g,- 10 | dB- W for g,—10? | gg: w for gi 107
323 2.1023 | 242 15-105 | 963 2.107
333 2-103 | 252 15-10'* | 273 2- 1018
Pigi Pi 3 Pigs P "
dB: W for &1=!0 [ag.w for gi !0
33 2-10! 152 108
215 3-10" | 134 2.5 - 108
197 6.10 |. 116 4 «10? Š
184 2. 10? 103 4-10,
169 8-107 88 6 +107)
158 7.109 77 6 -10

Detection; 8-410; t2=100 sec

directional reception,
s,-10*m*; r, -10*K

1.5. 1015
1.5 - 10/9

Pun Pi $
dB: W for g;=10

182 1.5 - 10°
192 1.5-10!°

Once two civilizations have discovered each other, they may establish

bilateral directional communication.* In this case, information can be

transmitted at a rate of 1 — 10? bits/sec over distances of 1000 light years
with power of the order of 109 — 10 watt, i.e., substantially less than the

power required for signal detection.

This, apparently paradoxical,

conclusion is quite understandable: the high power needed for signal
detection is the price we have to pay for not knowing the subscriber's
address.

* The concept of bilaterally directional radio communication does not refer to a dialogue between
civilizations (no such dialogue is possible in interstellar communication, since the answer will take
thousands of years to cross the tremendous distance), but rather the two unilateral "monologues" trans-
mitted through a channel with a directional transmitting antenna and a directional receiving antenna.

115

EXTRATERRESTRIAL CIVILIZA TIONS

TABLE 3.5. Minimum power needed for signal detection and reception of information over distances of
1000 light years (Af, = Ma = Af; 4 —8 cm)

Reception of information

a~100 Detection; p=10; t,=100 sec

Transmission

nondirectional re~ |directionalreception,

nondirectional re- Hirectional reception,
ception, 7, -10*K | S; «10'm?; r, =10°K

ception, 7, =10°K |s:-10'm?^; r,-10?

] |Isotropic
continuous Af =
1 2.10?! 15-1019
2-107 1.5. 10!
2.1023 1.5.10?
2.107 1.5. 1018
2.1025 1.5.10”
2
3 | Directional Pigi 1 $ P, Pig, P,
continuous Af=| dB- W for g:= 10 : dB: W for gi l0 | dB-W for gi — 10?
I 233 2.105 152 5-105 213 2.1012 132 1.5. 10*
10? 253 . 2.10” 142 1.5- 10%
2.101 152 1.5- 108
2. 1085 162 1.5- 107
2.106 172 1.5- 10°

Let us now try to establish the cost of interstellar communication.
Consider two civilizations at a distance of 1000 light years from each
other which have established bilaterally directional radio communication
using a high-frequency channel of 1 Hz band width. From Table 3.5 we
find the transmitter power needed for this communication; it is 1000 kW,
which is easily accessible to civilizations on our level. The signal-to-
noise ratio at the reception point will then be 100, and this is again quite
adequate for establishing reliable radio communication with the aid of
binary PCM. The transmission rate of PCM information through a 1 Hz
wide channel is 1 bit/sec. Let us find the cost of a 100 word transmission
through this channel. The message is composed using a 30-letter al-
phabet and each word contains on the average five letters. Our message
then contains 2.5 * 10? bits of information (see $ 1) and if transmitted at
a rate of 1 bit/sec, it will take 2.5-10? sec. Assuming a transmitter power
of 1000 kW, the message will consume about 700 kW-hr of energy, costing
about 28 rubles. We see that the communication with extraterrestrial
civilizations is not very expensive. The main problem is to discover your
counterpart.

116

RADIO COMMUNICATION WITH EXTRA 1 ERRESTRIAL CIVILIZATIC ^.

HI.

OkMOI- "4 'OT- d 'oor-?

‘WD G= Y IO] ƏPEWU 919^ suone[no[eo ayy 'sesmd ay) uso^1oq duroeds oum eun sı ; :uondooa: [euorj201rp 's[eugrs os[nd
JO uorssrursuef [euonoairp — 6 :uondooo: pue uorssrursuen [EUOTIDITP 10} uoro1ap Jo aBueI — g 1,01 -g vol "3
'uondooa1 [euoroa1Ip pue uorsspuisuer [EUOT221ID 10} uorie2Tunurui02 jo aZuei — 1, tuondaoo: [euorj2oirp 'speugrs os[nd
Jo uorsstuisuen ordonos: — 9 *uoridooo1 p[euorioaip pue uogsspusuen ordonost Joy uot12212p jo adueI — e ner =S
?uueaiue ZUTATIOOI JO Poi? IANOIJJƏ ‘uotdadal [euorio21rp pue uorsstuisue1) o1donos: 10} uonvorunuiuroo Jo oSue1

— y !uondeoosr jeuonooirpuou 'speuSis os[nd jo uorsstuisue1i otdonost — g !uondooo: [euorioo1puou pue uorssiusuen
atdonost 10} uon2o3ap Jo a8uer — e, tuondo»e1 peuorioairpuou pue uorsstuisuen o1donos: 10J uoneorunuiioo jo 23ue1 — T

"uorssrurue1 pue uoridoooi Jo spop w
SONA IOJ Sq;dua[aAe^ 10Jaurriuao Je aBueL uoneorunuturo3 “SA ( {v= IN = 4{¥) samod rentuisuen waunuyy "ey TURD

10 105070 10 0 io "o o"

* 3 8 a E R ARR
SA ` Sh à 85 h N “RRR

M “mod

117

EXTRATERRESTRIAL CIVILIZATIONS

Figure 45 plots the minimum transmission power vs. the range of
communication for various methods of reception and transmission. In
the top left corner of the figure lies the region of isotropic transmission
of wide-band signals with nondirectional reception. In the bottom right
corner we have the region corresponding to directional transmission of
narrow-band or pulse signals and directional reception. All the inter-
mediate cases fall in between.

It is left to the reader to choose on the ordinate axis the power values
corresponding to his skepticism or imagination, and to determine the
optimum methods of transmission and reception for any distance in the
Universe.

Let us try to estimate the possibility of detection of EC which do not
send special signals (by "listening in" to their internal radio transmissions).
The power involved in these transmissions will be of the order of 105 — 108
watt. Atthis power level, the isotropic wide-band ultra-short-wave
transmissions cannot be detected even from the nearest stars. Interception
of highly directional radio transmissions with a directivity coefficient of the
order of 10°, which extraterrestrial civilizations may use for some special
purposes (e.g., interplanetary communication), is possible over distances
of a few hundreds and even thousands of light years. However, the pro-
bability that such a tight message will be accidentally intercepted by the
receiving antenna is very low. The planetary rotation increases this
probability, but it nevertheless remains low, if we remember that at a
distance of 100 — 1000 light years there are less than ten transmitting
EC. Thus detection of extraterrestrial civilizations with
the aid of their routine radio communications is
virtually impossible. To become detectable, they
must transmit special signals in the form of powerful
isotropic radiation or in the form of highly directional
radiation with a scanning antenna.*

Radio communication between galaxies

Let us consider some specific features of intergalactic radio com-
munication. There is probably at leást one civilization per galaxy capable
of transmitting and receiving information, so that it is worth trying to
probe the individual galaxies with directional receiving and transmitting
antennas. The antenna directivity should be chosen so that the beam
would cover the entire galaxy, i.e., so that p=6, where ọ is the angular
size of the galaxy, and 0 is the beam width. Using relation (3.55)
between beam width and directivity coefficient and the dependence of the
angular size of a galaxy on its distance, we may write this condition

in the form
Ry? (3.69)

where L is the mean linear size of a galaxy. Suppose a civilization
situated in a certain galaxy sends an isotropic transmission (g,= 1) and

* From the point of view of reception and power requirements, this system is equivalent to isotropic
pulse transmission.

118

III. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

another civilization in a nearby galaxy uses a directional receiving antenna
of gain g, satisfying (3.69). Inserting the corresponding values of gi and go
in (3.33c), we obtain an expression of the power needed for this com-
munication:

Pi = 16akT, M (I). (3.70)

This is a paradoxical result, because the required power turns out
to be independent of distance! The same conclusion is obtained if
some EC sends a transmission directed at another galaxy with an antenna

2
which completely covers the recipient (ei =10 i. whereas the civilization

in the receiving galaxy uses a nondirectional antenna (g,=1). Let us
calculate the numerical value of the power needed for this kind of inter-
galactic communication. Setting in (3.70) a= 100, T,=10°K, 1—3 cm,
Af=1 Hz, we find P,—2*10?8 watt.

In case of bilaterally directional communication between the galaxies,
the receiving and the transmitting antenna gains satisfy (3.69). The power
needed for this communication is therefore

l6akTn AfL‘
PU. RS (3.71)

This result is even more puzzling: the greater the distance
between the communicating galaxies, the lower is the
power needed for establishing the communication!
Troitskii /6/ was the first to call attention to these peculiar features
of intergalactic communication. They seem to stem from condition
(3.69), i.e., they are basically associated with the fact that the antenna
directivity is a function of the angular dimensions of the galaxy (the area
of each antenna has to be increased in proportion to the intergalactic
distance). The antenna parameters and the power required for inter-
galactic communication emerge from Table 3.6, which lists the directivity
coefficients, the receiving antenna areas, and the power for communication
with two neighbor galaxies (the Large Magellanic Cloud and the Andromeda
Nebula) and with some other typical galaxy with dimensions of the order
of 10? light years. Here, as before, «— 100, Ta =10°K, Af Af -A[A1 Hz.

TABLE 3.6. The directivity coefficient of antennas and the required power for intergalactic communication

Large
Galaxies Magellanic| Andromeda A typical galaxy with L « 10 light years
Cloud
Distance in light years . 2-108 2-108 10! 10° 10° 10%
Angular dimensions ... 9? 3*5 34' 206" 21" 2"
Directivity coefficient of
receiving antenna ... 360 3-10° 108 107 10? 10"!
Area (mô at 4 — 3.5cm . | 0.04 0.3 10 10° 10 107
Transmitter power, watt. | 6-10? 104 2:10? 2.10? 2-10! 21016

119

EXTRATERRESTRIAL CIVILIZATIONS
Monochromatic signals. Frequency scanning

We have so far considered the power aspects of communication. Now
we can discuss in more detail the various aspects relating to signal band
width. Here we should distinguish between two cases: wide-band signals
with a virtually continuous spectrum (Av~v) and narrow-band monochro-
matic signals with a band width substantially narrower than the frequency
(Av«v).

Frequency scanning constitutes one of the basic stages in any search
for narrow-band signals. As we have noted before, the band width in low-
speed transmission (e.g., in case of call signals) is determined entirely
by the stability of the transmitted signal. It may be as low as fractions
of Hz. On the other hand, the width of the optimum frequency range where
EC signals can be expected is of the order of 10 — 10!! Hz. Our problem
is thus how to detect a narrow line of relative width of 10-1! — 107" in this
frequency range. Since the direction in which these signals should be sought
is not known either, the problem, to borrow Purcell's expression, is not
unlike that of trying to meet a certain person in New York City without
having previously agreed on a meeting place. Nevertheless, this complex
problem is basically and technically solvable.

We will first consider the question of frequency scanning, and then
proceed to direction scanning. There are two methods of frequency
Scanning currently known: a single-channel scanning receiver with
automatic frequency tuning or a multichannel receiver with narrow-band
filters, each tuned to a certain frequency and all the filters jointly covering
the entire frequency range. Which of the two techniques is to be preferred?
To answer this question, we have to use some logical evaluation criterion.
Siforov /8/ proposed the following criterion. In order to detect EC signals,
we have to ensure reception of a sufficient quantity of information which
will indicate with high reliability the artificial origin of the radio source.

It is therefore best to use those signal detection methods which provide
the essential minimum of information in minimum time. Let us now
evaluate the two frequency scanning techniques from this point of view.

Consider a single-channel scanning receiver with continuous frequency
tuning, the block diagram of which is shown in Figure 46.

FIGURE 46. Block diagram of a single-channel frequency -scan-
ning receiver:

A — antenna, HFA — high frequency amplifier, O — oscillator,
M — mixer, F — filter, IFA — intermediate-frequency amplifier,
D — detector, LFA — low-frequency amplifier, R — recorder.

120

IIl. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

The oscillator frequency v4, follows the frequency of the tunable HF
amplifier. The mixer transforms the instantaneous high frequency v into
a constant intermediate frequency vig v— v, After the mixer, the signal
is passed through an intermediate-frequency filter, amplified, detected,
and transmitted to a recorder through a low-frequency amplifier. The
receiver band width Af, is limited by the intermediate-frequency filter,
Afe=Afine Let At be the time for the current frequency to scan a frequency
band Afe. As this band width is being scanned, the filter input receives

a certain signal x(/) of duration Af whose band width is Af=Ai-!. To ensure
undistorted transmission of this signal through the filter, we should clearly
have Af<Af, or At>Af;'. Hence it follows, that the rate of frequency
variation in this tunable system depends on the receiver band width Af;:

a ar NAR). (3.72)

This rate of tuning is limited from above, and the time of scanning of a
given frequency band therefore cannot be made as small as desired. The
maximum rate of frequency variation is

(2)... E: (AY (9:19)

and the corresponding minimum scanning time for a frequency band Af, is

Aus = AR". (3.74)

min

Let Afo be the scanned frequency range. The total scanning time
required for the current tunable frequency to run through the entire
relevant range is thus

^
Tint ^N Almin Mis , (3.75)

where N is the number of elementary frequency bands, i.e., receiver
bands Afe, accommodated in the scanned frequency range. We thus see
that the time of frequency scanning is inversely proportional to the square of
the band width of a single-channel scanning receiver. This is quite under-
Standable, since a decrease of the receiver band width, on the one hand,
increases the number of elementary bands into which the scanned frequency
range is divided and, on the other, increases the scanning time in each
elementary band (since the rate of frequency variation decreases in
proportion to the square of the band width). .

Let us now express the scanning time as a function of the distance to the
source. Since Af;«R-?, equation (3.75) may be written in the form

TOC MoRÁ. (3.76)

Thus, for the particular reception technique using a single-channel
Scanning receiver, the time of search for monochromatic EC signals is
Seen to be directly proportional to the total frequency band Afo in which the
Search is conducted and to the fourth power of the distance to the sending
civilization.

121

EXTRATERRESTRIAL CIVILIZATIONS

If we use a multichannel receiver made up of N channels of band width
Af» each, all the channels fully covering the required frequency range
(NAfo=Afo), the total frequency scanning time will be equal to the time to scan
a single channel. It follows from (3.74) that in this case the time is in-
versely proportional to the band width of each channel or, using the relation-
ship between range and band width, it is proportional to the square of the
distance between civilizations.

Consider a search for signals at 3 cm wavelength when the distance
to the sending civilization is 1000 light years, the transmitter power is
150 MW, the transmitting antenna gain is 90 dB, the effective area of the
receiving antenna is 10* m?, and the noise temperature is 10*K. Suppose
that information can be received for signal-to-noise ratios of 100. We
See from Table 3.5 that the receiver band width in this case should be
100 Hz. The minimum scanning time for this band width is of the order
of 0.01 sec. Ifa single-channel scanning receiver is used, several years
will be needed to scan the entire frequency range around 3 cm (Afi 10? Hz),
A multichannel receiver comprising 10? channels each 100 Hz high will scan
the entire frequency range in a time of the order of 0.01 sec. If the
distance to the sending civilization is increased 10-fold, the scanning time
with a multichannel receiver will increase 100-fold reaching 1 sec, whereas
for a single-channel scanning receiver the time will increase by a factor
of 10*, reaching 30,000 years.

Wethusseethat the scanning time of a continuously tunable
single-channel receiver is much longer than the scan-
ning time of a multichannel receiver. Moreover, as the
distance between the civilizations increases, the scanning time witha
single-channel receiver grows much faster (1,«R‘) than the scanning time
of a multichannel system (x, R2).

All this renders single-channel scanning receivers practically useless
for the detection of monochromatic signals from extraterrestrial civiliza-
tions. This problem can be tackled more successfully using multichannel
receivers with a great number of narrow-band filters. It should be kept in
mind that the reduction of scanning time is accomplished as a result of
a much greater complexity of the receiving equipment, the instrumental
complexity (the number of channels jn the receiver) increasing in proportion
to R?, Nevertheless, it seems that this complexity is not without its
advantages /8/, since band filters (and other components used in multi-
channel systems) are cheap and readily accessible.

A multichannel system specifically designed for the detection of mono-
chromatic EC signals was proposed by Kotel'nikov /16/. A block diagram
of the receiver is shown in Figure47. Here Ais theantenna, AMisthe amplifier
which also transforms the frequency of the incoming signals, ifnecessary, Fare
filters of bandwidth Af, covering jointly the entire frequency range, D are the
detectors, Iare the integrators integrating the energy which passes through the
filter inatimet, Tare the threshold devices which produce an output signal only
if the energy transmitted through the filter inatime t exceedsa certain
threshold value.

This receiver clearly cannot be used in reception of information
transmitted by one of the amplitude modulation techniques. However,
information may be sent by varying the frequency of the signal from one
transmission to the next. In this case, a signal will appear in one of the
receiver channels, and every successive transmission will be picked up
by a different channel.

5780 122

III, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

FIGURE 47. Kotel'nikov's multichannel frequency-
scanning receiver:

A — antenna, AM — amplifier, F — narrow-band filters,
D — detectors, 1 — integrators, T — threshold elements.

The appearance of a signal in a new channel may be regarded as a
certain message. Since one out of N possible channels is selected, each
of these messages contains log, N information units (see $1). The rate
of information transmission of this system is thus

q= DRM. (3.77)

Let us now find the range of communication. For bilaterally directional
radio communication, the range of signal reception with Kotel'nikov's
receiver is

R=( gare) (3.78) 228
The factor V in the denominator depends on the number of channels N,

the band width Af of each channel, and the particular threshold used,

which in its turn is determined by the probability of a false response

P. of the system (as a result of random noise in the receiver) and

the probability of missing a signal pms- The range of communication

increases with the decrease in V. The minimum value of V is attained

for M-l; it is equal to

N I 2
Y t V nne. aa 3.79
min ( l Pec. Ins 2) . ( )

miss

This V for a given t gives the maximum range of communication:

= Pig. S: Ya
Roo = (woes) (3.80)

This expression is analogous to (3.33). The only difference is that the
a in the numerator has been replaced with Ymm; its meaning is the same
as that of a, since it is determined by the system threshold. It follows
from the last two relations that the maximum range of communication
falls with an increase in the number of channels and an increase in the

123

EXTRATERRESTRIAL CIVILIZATIONS

band width of each channel. If a certain frequency band Afo is to be scanned,

we have Af= Ale, As N increases, the frequency band of each channel

decreases faster than Wgi, increases; as a result, the maximum range of
communication increases with the increase in the number of channels. For

A 4, we have
V = (2 mint Af)” , (3.81)

and for a given frequency interval Afo, the range of communication is

{Pags tN Ys
R=( RE V wea) (3.82)

Let us consider one example. An EC transmitter of 100 MW power
sends monochromatic signals in the 3 cm wavelength range in the form
of pulses of 100 sec duration, varying the pulse frequency from one trans-
mission to the next. The transmitting antenna has a gain of 10°, and the
reception is carried out with an antenna of 10* m? area and Kotel'nikov's
receiver with M=10° channels, noise temperature 7, =10°K, integration
time 1— 100 sec, channel width 1 Hz; the false response probability and
the signal omission probability are 1075, Inserting these values of
N, v, Af, Pi and p, in(3.79) and (3.81), we find w—120. Equation (3.78)
then gives the range R=7-10* light years. The transmission rate of this
communication system is 0.01 log, 10? = 0.3 bit/sec. As t decreases,
the range of communication slowly diminishes, whereas the transmission
rate increases fairly rapidly. If we take in our example t=1 sec, we
find Y=Y¥mn=73, R= 9-10? light years, and the transmission rate will
increase to 30 bit/sec.

It is interesting to compare these numbers with the corresponding data
for a single-channel receiver operating at a fixed frequency. Using
Figure 45 we find that for P,=108 watt, gi— 109, S,=104 m?, 7, =10°K,
Af=1 Hz, and «— 100 the range of bilaterally directed communication is
8-10? light years and the transmission rate is 1 bit/sec, i.e., 1/30 of
the transmission rate attainable with Kotel'nikov's system for the same
range of communication. The range of detection for these system
parameters and B=10, te=100 sec is 8- 10* light years, i.e., of the same
order of magnitude as the range of communication attainable with
a multichannel receiver with transmission rate of 0.3 bit/sec and equal
other parameters.

We thus conclude that Kotel'nikov's multichannel receiver
is an optimal system for searching for monochromatic
signals when the frequency band to be scanned is not toc
wide. However, the number of channels required to detect a line
narrower than 1 Hz in a frequency band of 10!? — 10!! Hz is uncomfortably
large. To avoid this difficulty, Troitskii proposed an original combination
method. A special spectrum analyzer is applied to cover simultaneously
a sufficiently wide part of the spectrum with band width Afo of the order
of 1 MHz. In this way, the presence or absence of a monochromatic
sine signal in some frequency range can be immediately established. The
exact frequency of the signal is not determined, and only some wide

124

II, RADIO COMMUNICATION WITH EXTRATFRRESTRIAL CIVILIZATIONS

frequency band Af» containing the signal is identified. Once the relevant
frequency interval has been identified, a multichannel receiver is applied
to exactly determine the signal frequency. For a channel width of 1 Hz,
109 channels are required to cover a band of Afo=1 MHz in which a signal
has been detected. This is not an excessively large number of channels.
Moreover, the construction of a multichannel receiver covering the par-
ticular frequency interval will be justified by the detection of a signal in
that interval.

This method should be first applied to frequencies
near the 21 cm hydrogen line, near its harmonics, near
the 18 cm hydroxyl OH line, andalso possibly near the1.25 cm
ammonia line and the 0.4 cm formaldehyde line used in molecular masers.

Direction scanning

Let us now consider the direction scanning in the general search for
signals. Suppose that the distance to the nearest civilization sending
meaningful signals into space does not exceed some value R. Then the
signals can be detected by examining the stars lying in a sphere of radius
R around the Sun. How many stars will have to be examined in this way?
The mean interstellar distance in the neighborhood of the Sun is about
2.2 pc (i.e., about 7 light years). The stellar density here is thus 0.1
stars per pc? or 0.003 stars per cubic light year. Let R— 1000 light
years. A sphere of this radius will contain 10 million stars. The number
of suitable candidates can be reduced if we remember that only a small
fraction (at most 1%) may have planetary systems capable of sustaining
life. We are thus faced with a very difficult and highly undeterminate
problem: to choose a few hundred thousand stars from among 10 million
which are likely to sustain highly developed civilizations. Ironically,
the situation is much simpler with the search for civilizations in other
galaxies (this problem has been treated above). Let us return to stars,
however.

An optimum system for a search for signals sent from an unknown
direction comprises directional fixed antennas whose beams cover the
entire sky. If the sending EC transmits in all directions in space, it can
be detected without difficulty. We have seen, however (Table 3.5), that
this isotropic transmission requires a tremendous transmitter power and
a highly directional receiving antenna. For a distance of 1000 light years
and a transmitter band width of 1000 MHz, a power of the order of 10?*
watt is required (this power is available only to type II civilizations)
and the receiving antenna should have an effective area of 10* m? (assuming
T,=10°K). This antenna has a directivity of 10? in the centimeter range,
and some 100 million such antennas will be needed to cover the entire sky.
Such a detection system clearly may be set up within the next 100 years.
However, this project falls beyond the current financial resources of man-
kind.

The requirements regarding antenna area, the number of receiving
antennas, and transmitter power can be relaxed if a directional trans-
mitting antenna is used. Following V. A. Kotel'nikov, let us consider
two civilizations A and B, distant R from each other. Civilization A
transmits with a highly beamed antenna, and civilization B is at the

125

EXTRATERRESTRIAL CIVILIZATIONS

receiving end. Civilization A is not aware of the location of civilization

B and the direction in which the signals should be sent is not known to start
with. The antenna beam should therefore "trace" the entire Sky. Let t

be the transmission length and «o the antenna solid angle (o € ). To scan

the entire celestial sphere, it takes

f= = r= ga. (3.83)

Suppose that civilization B has a detection system which comprises
an assembly of directional antennas covering the celestial sphere. One
of these antennas is aimed at civilization A. The receiver connected to
this antenna records a signal at the time when the transmitting antenna
of civilization A is aimed at civilization B. The signal detection experi-
ment will take a time ¢ much longer than fy. The signal from civilization A
will therefore be picked up several times, at equal time intervals 4%.

In this way, it can be reliably distinguished from random noise. The
time of detection can be somewhat reduced if civilization A, instead of
Scanning the entire sky, will concentrate on a limited number of suitable
stars lying in a sphere of radius R (which naturally includes civilization B)
and then send signals only in the direction of these chosen stars, shifting
the antenna from one star to another.

Suppose that civilization Ais distant 1000 light years, the transmitter
power is 10!7 watt (this is available to type I civilizations), A/- 1000 MHz,
transmitting antenna directivity g—10?, 2=3.5 cm, transmission time
t=3 sec. From (3.83) we then find that 3- 10? sec or 100 years will be
needed to scan the entire celestial sphere. The time to scan all the
stars inside a sphere of 1000 light years radius is 3-107" sec = 1 year.

If only the most suitable stars are scanned (assuming that about 176 of the
stars will support advanced civilizations), the total scanning time will be
3:10? se» or about 3.5 days. With a transmitter power of 10" watt and

a band width of 1000 MHz, a fairly humble receiving antenna with an
effective area of about 100 m? (and T, = 10*K) is required to detect signals
over a distance of 1000 light years. One million such antennas will cover
the entire sky simultaneously. Kotel'nikov proposed using multibeam
antennas (technically, this is feasible, since each antenna is stationary),
and the number of covering antennas can be reduced at least by one order
of magnitude in this way. To reduce the number of antennas even further,
he suggests dividing the sky into several areas and studying each area
separately. This naturally will lengthen the detection time. Thus, in our
case, when the scanning time for all the suitable stars is about 3.5 days,
the celestial sphere can be divided into 10 parts, scanning each area

in 36 days (during this period, the signal should appear at least ten times);
the entire experiment will then be completed in 1 year, and it will require
104 ten-beam antennas. If R=100 light years, the scanning time for the
suitable stars is 300 sec. The sky can be divided into 10* areas and, scan-
ning each area for 3000 sec (the signal will appear at least 10 times during
this period), we will complete the experiment again in 1 year. The re-
ceiving antenna area needed to detect signals over a distance of 100 light
years is 1 m?, and the antenna directivity is 104, i.e., a single antenna
will cover each of the areas! If R=104 light years, the scanning time for

126

HL RADIO COMMUNICAIION WITH EX [RALFRRESTRIAL CIVILIZATIONS

the most suitable stars will be 3-108 sec or about 10 years. Further
division of the sky into areas will greatly prolong the experiment. And
yet, for complete coverage of the sky at this distance, we will need 10?
antennas of 10* m? each. Thus, given the transmitter power
and bandwidth, the directivity of the transmitting
antenna, and the transmission time t, we can establish
the optimum distance to other civilizations for which
multiantenna detection systems are practicable. In our
example, this optimum distance is of the order of
1000 light years.

The above reasoning applies to the search for wide-band signals, as
well as monochromatic signals. The only difference is that ina search
for monochromatic signals each antenna of the detection system should be
provided with a multichannel receiver. For example, consider Kotel'nikov's
receiver with 10? channels, 7, =30°K, Af=0.3 Hz; let the transmitting
civilization use a 10? watt transmitter and a 109 m? antenna, sending 10 cm
monochromatic signals of duration r=3 sec. Signal detection will then
require a multiantenna system whose parameters are listed in Table 3.7.

TABLE 3.7. The parancters of a detection system for monocliomatic 1C signals (P, = 10° watt, S, = 109m,
R= 1an, t=3ece, N=10% 5/5 20,3 Hz, T, =30°K) according to Kotel nikov /16/

5 = Zo zn LL
z a S Due E " S
z z% z z 3 E "EA EET
p ki É EH per 8 b E z sy yz č i
= E E Er z gu s 8c Ee
z <2 š zi = z 5 29 eum m ee 42
= a = -e = = -

2 = Ep els ae cae ks) 2 es Je"
g Paes s E rS EE ou 7 Ze: zz
S & £g ES EX Tu $8z| 274 2B
z 5g By Grecs p gat Egg Eas
P = [a dj m 2% 3 a 2 5 3 2c Se 9 gre
2 “ T pu F Fa aad Zor AA zd
2000 10° 10 years 26 days 400 480 900 1 480 020
1000 10' I y.ar 4 days 100 120.000 10 12.000

" ) )
20 10 36 day. hones 25 30 000 109 300
200 10° 4 days i hour 4 4.302 1020 5

On the basis of these data, Kotel'nikov came to the conclusion that
radio signals from civilizations of our (or slightly higher) level definitely
can be detected if there is at least one such civilization in 10° stars. If
there is only one civilization in 107 stars, its detection presents a much
more difficult problem, but it is nevertheless feasible under certain
conditions. One civilization in 10? stars is extremely difficult to detect
by the present-day means.

Wide-band signals. Sky surveys

Consider the search for wide-band signals. When the band width is
of the order of magnitude of the transmitted frequency, the artificial
signal is similar to the radio emission of natural sources. This leads
*o two conclusions. First, wide-band signals can be detected using

EXTRATERRESTRIAL CIVILIZA TIONS

conventional radio astronomical equipment. Second, to detect wide-band
signals, we should first establish how to distinguish the artificial from
natural signals. After all, before attempting to decode the signal, we must
be sure that we are dealing with an artificial source, which has to be
identified among a multitude of natural radio sources. This brings us back
to the problem of artificiality criteria, discussed in §3.

Any systematic search for artificial sources should include as a first
step the discovery of all the radio sources followed by sifting in accordance
with the likely artificiality criteria. Complete sky surveys in the radio
spectrum should thus be launched. The meter and the decimeter wave-
length range has been studied in fairly great detail. Detailed catalogues
have been assembled for these wavelengths, listing all sources with
radio fluxes down to 10778 watt/m? . Hz. The situation is much worse
in the centimeter range, however. No complete sky survey has been
carried out in this range, and yet it is at these wavelengths that the
civilizations are likely to communicate. Therefore, one of the
immediate tasks is the organization of a detailed sky
Survey in the centimeter range using high-sensitivity
astronomical equipment. i

What are the requirements to be met by a radio telescope used in this
survey? We should naturally strive to minimize the total survey time.

And yet the radio telescope should have a maximum sensitivity or, in

other words, the receiving antenna should be made as large as possible.
Sky surveys can conveniently utilize the diurnal rotation of the earth.
Consider a radio telescope with the antenna axis fixed in the meridional
plane. The diurnal rotation of the Earth will successively aim the

antenna pattern at different areas of the celestial sphere, all lying on

the same diurnal parallel. In 24 hours, the telescope will survey a ring
strip of the sky of width 20,, where 20, is the vertical width of the antenna
pattern between half-power points. Now we can displace the antenna through
a distance 20, in declination, and it will survey a new annular strip during
the next day; this strip adjoins the previous one and has the same width 20,.
The total time to survey the entire sky will clearly be

To— 35, ~ a days, (3.84)

where h is the vertical dimension of the radio telescope dish. The survey
time thus decreases as 0, increases or as the vertical dimension kh of the
dish decreases. For a given surface area, the minimum
survey time is ensured if the vertical dimension of the
reflecting surface is much less than the horizontal
dimension, while the vertical dimension of the antenna
pattern is much greater than the horizontal dimension
(6:250) In other words, the radio telescope should have
a "knife-edge" antenna.

Consider a radio telescope with an antenna in the form of a paraboloid
of revolution 50 m in diameter (surface area 2000 m?). This antenna has
a symmetric pattern, whose width at 1 cm wavelength is 20,=20,=28) ;,22*10*
radian = 40". Inserting this value of 0, in (3.84), we obtain for the total
survey time To= 43 years. Let us now consider an antenna in the form of a
parabolic cylinder with horizontal span /[— 400 m and height k=5 m.

128

Il, RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

This antenna, for the same geometrical area of 2000 m?, has a "knife-edge"'

pattern with 20,2 5" and 20,— 7! (at 1 wavelength), and the total survey time
will be 4.3 years. We see from this example that a radio telescope with a
"knife-edge" antenna pattern not only greatly reduces the survey time, but
also ensures a high resolving power (at least in one coordinate). When
considering very large radio telescopes, whose size approaches the limit
fixed by effects associated with radio brightness fluctuations of the meta-
galactic background and the atmosphere, we notice another important
advantage of "knife-edge'" antennas: they ensure the maximum sensitivity
for a given antenna surface.

Two "knife-edge'' antenna designs are currently known: Kaidanovskii
and Khaikin's variable profile antenna (VPA) and the Krauss radio telescope.
Figure 48 is a photograph of the large Pulkovo radio telescope with a
variable profile antenna. The telescope is made up of separate shields,
mounted along the arc of a circle. Each shield can be moved along the
circle radius, turning in azimuth and position angle. By appropriately
moving the separate shields, the reflecting surface can be rearranged
so that the radio telescope is aimed at a desired point of the sky. The horizontal
width of the antenna pattern is determined by the horizontal span of the antenna
(the length of the chord spanned by the working sector); the vertical width for
observations near the horizon is determined by the height ofthe shields. As the
position angle increases, the vertical width of the antenna pattern diminishes,
and in the zenith (when the VPA is a closed circle) it is equal to the horizontal
width: the "knife-edge" pattern is thus transformed into a ''pencil-beam" pat-
tern. This effect increases the survey time. Krauss' radio telescope is more
suitable for sky survey purposes (Figure 49). It consists of two separate
reflecting surfaces: a fixed parabolic reflector whose optical axis is aligned
in the meridional plane, and a moving plane reflector which may be rotated
about a horizontal axis, ensuring observations at various position angles
in the meridian. This radio telescope has a 'knife-edge' pattern, whose
vertical width is determined by the height of the parabolic reflector and
is independent of the source position angle. A slightly modified form of
this radio telescope, operating at 21 cm, has been recently built in France
(Figure 50).

FIGURE 48. The large Pulkovo radio telescope with variable profile antenna.

129

EXTRATERRESTRIAL CIVILIZATIONS

FIGURE 49. The Krauss radio telescope (USA).

When the sky survey has been completed, the sources should be sorted
out according to some criteria. The criterion of angular dimensions is
probably the most suitable to this end. One of the possible approaches is to
identify all the sources with angular dimensions less than 0"'.1 and radio
fluxes up to 10-2? watt/m? - Hz (omitting all the natural sources of large
angular dimensions). This problem can be solved with high-sensitivity
radio interferometers consisting of large antennas of 10? — 10* m? areas
Separated to a distance of the order of 109 — 107 wavelengths (in the centi-
meter range).

The selection of radio sources with angular dimensions of less than
0'.1 should be regarded as the first preliminary stage of the program.
Radio interferometers with an ultralong basc, using the existing network
of radio telescopes (a global system of radio interferometers), will attain
a resolving power of 0".001 (a resolving power of 0'.005 has already been
attained). In the future, Earth spacecraft radio interferometers will probably
be created. This will ensure bases of the order of 1 a.u. and reach
resolving powers of the order of 10-7? angular second in the centimeter
wavelength range. The selected sources will then have to be carefully
studied using the various artificiality criteria. This opens wide horizons
for future studies.

From the point of view of radio astronomy, artificial radio sources must
possess certain unusual properties, i.e., an artificial source is a priori a
peculiar radio source. The problem of discovering and studying peculiar
radio sources is one of the basic tasks of radio astronomy. In this respect,
our problem of search for extraterrestrial civilizations is closely linked
with one of the most topical and pressing problems of radio astronomy.

130

HL RADIO COMMUNICATION WITH EXTRA [FRRES TRIAL CIVILIZA TLONS

IE 31y 8:943 Baa dels Ve
DOR ff Yan as.

e AUS

> aio Er ae
WOE top ste nno v.

19. ban Fi ss wan nm Sd
$5

4 ov

FIGURE 50. The Nangay radio telescope (France), operating at 21 cm wavelength,

The cffective antenna area is 7000 mô, horizontal heam width 355, vertical beam width 20'. The receiver has
15 channels of 280 kTIz band width cach.

Bibliography

1. Cocconi,G. and P. Morrison.-— Nature, Vol. 184:844. 1959.

2. Cameron, A. (Editor). Interstellar Communication. — New York.
Benjamin. 1963.

3. Vnezemnye tsivilizatsii (Extraterrestrial Civilizations). Proceedings
of a Conierence, Byurakan, 20—23 May, 1964.— Izd. AN Arm. SSR.
1965.*

4. Hartley,L.V.L. Transmission of Information. — BSTJ, 7 (3):
535—563. 1928.

5. Shannon,C.E. Communication in the Presence of Noise. — PIRE,

37 (1):10—21. 1949.

6. Troitskii,V.S. Nekotorye soobrazheniya o poiskakh razumnykh
signalov iz Vselennoi (Some Considerations on the Search for
Intelligent Signals from Space).— In: Extraterrestrial Civilizations,*
/3/, pp. 62—71.

7. Webb,J. Discovery of Intelligent Signals from Outer Space. —

In: Interstellar Communication /2/.

* [See footnote on p. 11.]

131

10.

11.

12.

13.

14.
15.

16.

EXTRATERRESTRIAL CIVILIZA TIONS

Siforov,V.I. Nekotorye voprosy poiska i analiza radioizlachenii ot
drugikh tsivilizatsii (Some Aspects of the Search for Radio Signals
from other Civilizations andtheir Analysis).— In: Extraterrestrial
Civilizations* /3/, pp.78—83.

Shklovskii,I.S. Izluchenie "misteriuma" kak lazernyi effekt
("Mysterium" Radiation as a Laser Effect). — Astr. Tsirk., No.
No.372:1—8. 1966. R

Kardashev,N.S. Peredacha informatsii vnezemnymi tsivilizatsiyami
(Information Transmission by Extraterrestrial Civilizations). —
Astron. Zhurnal, Vol.41 :282. 1964.

Slysh, V.I. Radioastronomicheskie kriterii iskusstvennosti radio-
istochnikov (Radio-astronomic Artificiality Criteria of Radio
Sources). — In: Extraterrestrial Civilizations* /3/, pp. 38—42.

Gudzenko, L.I. and B.N.Panovkin. K voprosu o prieme signalov
vnezemnoi tsivilizatsii (Reception of Signals Transmitted by
Extraterrestrial Civilizations). — In: Extraterrestrial Civilizations*
/3/, pp. 43—45.

Sholomitskii, G.B. Fluktuatsii potoka CTA-102 na volne 32,5 cm
(Flux Fluctuations of CTA-102 at 32.5 cm wavelength). — Astron.
Zhurnal, Vol. 42:673. 1965.

Golei, M. Coherence of Intelligent Signals. — In: Interstellar
Communication /2/.

Shklovskii,I.S. Vselennaya, zhizn', razum (Life and Intelligence
in the Universe). 2nd Ed.— "Nauka." 1965.

Kotel'nikov, V. A. Svyaz! s vnezemnymi tsivilizatsiyami v radio-
diapazone (Radio Communication with Extraterrestrial
Civilizations).— In: Extraterrestrial Civilizations* /3/, pp.72—77.

* [See footnote on p. 11.]

132

Chaptev IV

METHODS OF MESSAGE DECODING

$1. INTRODUCTION

The problem of signal decoding evidently occupies an important position
among the various topics relating to communication with interstellar
civilizations.

Every astronomer, analyzing the signals from various celestial objects,
uses his own decoding system in the interpretation of his observations.
However, the information discussed in connection with extraterrestrial
civilizations is not the kind of information confined to the particular source:
this information in principle reflects the structure of the Universe, including
the organization of a certain society of "intelligent beings," i.e., it covers
approximately the same scope as the "terrestrial" literature.

A characteristic feature of the problem of decoding of messages from
extraterrestrial civilizations is the virtually total lack of any prior informa-
tion or knowledge about these civilizations. We are thus faced essentially
with a problem of decoding an arbitrary text.

Until recently, the problem of decoding of arbitrary texts did not attract
particular attention in linguistics. Nevertheless, some decoding methods
are available, using a minimum of preliminary information about the text.
The general ideas underlying these decoding methods appear quite interesting,
and the experimental results are very promising. It is hoped that the
"extraterrestrial bias" will provide a strong stimulus to the development of
this direction in linguistics.

There is always a chance that some accidental development will help to
decode the message. It would seem that the messages from extraterrestrial
civilizations would be organized in such a way as to simplify their decoding
as far as possible. It is more prudent to assume, however, that the decoding
of these messages will present considerable difficulties, no smaller, say,
than the decoding of inscriptions in ancient lost languages. This approach is
particularly important in that it prepares us for a linguistic struggle with
extraterrestrial messages, and does not limit our task to mere detection.
For a professional linguist, the tackling of codes and ciphers is a highly
attractive occupation, which requires deep insight into the structure and the
nature of language.

Interstellar linguistics also presents another problem (besides decoding).
This is the problem of creating the most effective language for interstellar
communication. It is particularly attractive in that every linguist goes all
the way toward creating a certain consistent language, whereas there can
hardly be a man capable of developing a full range of decoding methods.

133

EXTRATERRESTRIAL CIVILIZATIONS

However, the topical interest of this problem lies elsewhere, since inter-
stellar communication cannot take the form of a dialogue. In the best case,

a response to a message will be received after several centuries. If, on

the other hand, extraterrestrial civilizations will take longer over responding
to a message than it takes us to crack their code (or will lose interest
altogether), interstellar communication will never progress beyond the realm
of science fiction.

Interstellar communication is apparently not unlike literary activity: the
messages are broadcast by the author civilization in all directions (just like
books sent to various libraries and bookstores); the sender does not
expect any response, just as the author never writes a book for the sake of
a review. The reward is the privilege of getting acquainted with messages
sent from other worlds.

Mankind will clearly make its first steps in interstellar society as a
reader, rather than a writer. The problem of message decoding is therefore
much more pressing than the problem of developing interstellar languages,
at least at the present stage.

The aim of this chapter is to acquaint the reader with new linguistic
methods of message decoding.

These methods are computer oriented and therefore basically reduce to
algorithms, sets of instructions for a computer. For the reader's conve-
nience, the algorithms are presented in generalized condensed form, with
omission of most of the insignificant details.

The aim of decoding methods is two-fold. In practice, they are designed
for cracking code messages. Theoretically, decoding algorithms present
definitions of the linguistic features that they recognize in the message. In
this respect, they are of particular interest to the professional linguist.

The main significance of algorithms from this theoretical point of view is
that they provide general methods of analysis, suitable for repeated applica-
tion. The generality of the algorithms imposes natural restrictions on the
intuition and the whim of the linguist.

This two-fold aim presents different requirements to be satisfied by the
algorithms; on the one hand, they should provide accurate results, and on
the other hand, they should be as free as possible from arbitrary features
and logical ambiguities. For example, we tried to avoid the use of "empiri-
cal" numerical constants. In cases when the arbitrary approach was inevi-
table, we tried to apply simple solutions. This includes the construction of
"estimate functions" of maximum simplicity. Occasionally, we reproduce
algorithms which are known to provide unsatisfactory results, because their
"scheme" may prove helpful in future work.

The reader will notice that the material presented in this chapter is of
uniform interest. We wanted to focus our attention on the "basic" algorithms
— the algorithm of identification of two groups of letters, the semantic algo-
rithm, the algorithm of search for the sentence graph, algorithms identifying
code sequences and morphemes, pattern decoding algorithms, and letter
comparison algorithms.

Some readers may think that algorithms identifying vowels and consonants
have only remote relation to interstellar communication problems. We want
to stress, however, that all the algorithms are amenable to a more general
interpretation (this point will be discussed in greater detail later on).

134

IV. MESSAGE DECODING

$2. THE CONCEPT OF A MESSAGE, ITS
INTELLIGIBILITY AND MEANINGFULNESS

Definition of message

The aim of the present section is to provide an exact formulation of the
basic concepts and problems encountered in decoding. It is generally
assumed (and rightly so) that decoders deal with messages which should
be understood and translated into a known language. When
decoding messages received from outer space, there is an important
preliminary Stage: it is necessary to establish whether or not the message
is intended for decoding (or is worth the effort). In other words, we have
to establish first that the message is meaningful. These concepts of
intelligibility and meaningfulness will be analyzed in this section.

A message is a system M of three sets: the alphabet (the set of letters)
A={a,), the set of positions L—(/), and the text T—(a//), or the product of the
Set of letters with the set of positions, i.e., the set of pairs of the form ailj,
where a;e&4A,l;eL.

In case of a general message, no restrictions are imposed on any of the
three sets; they may be either finite or infinite, mathematically they may
present groups, rings, spaces, etc. This is clearly a very general concept,
and for many practical purposes the concept of a message should be properly
restricted.

The set A is generally assumed to be finite or at least enumerable; a
metric or topology is defined on the set of positions. Finally T, the set of
text inclusions, is generally characterized as a one-to-many mapping of the
set of letters into the set of positions, i.e., to each position /, is assigned
a single letter a;, but any letter a; can be found in any number of positions
in the set L. ;

The last condition leads to the highly important concept of an "absolute
frequency" of a letter aj.

The absolute frequency of a letter a; (of q(a;)) is defined as the power of
the set {a,l,}, i.e., the set of all textual pairs containing the letter a;.

The metric of the set of positions can be extended to the text: a textual
distance between the pairs a;l and a,l is naturally defined as the distance
between /, and 4.

Distance in the set of positions can be defined in a different way also. For
example, we can define the relation of adjacency by specifying what pairs
are adjacent and what are not.

Let A-(aj be the Russian-language alphabet. L consists of two ring
sections. The rings can be moved at random one relative to the other, and
the text appears as shown in Figure 51.

In this case, we cannot define distance between positions on two different
rings, but for each position we can identify two adjacent positions (on the
same ring).

This definition of a message may look too general. Why not define a
message in the usual way, as a string of letters?

There are examples, however, which make the conventional concepts
look quite unnatural. A drawing may not be considered as a message; on

135

EXTRATERRESTRIAL CIVILIZATIONS

the other hand, a linear scan of the same drawing is a message. The
definition of a text in terms of mapping into a graph suffers from similar

shortcomings. *
Nevertheless, it is desirable to formulate less general definitions of a
message for particular uses.
A message in a general sense will thus be characterized by the following
additional features: 1. The sets A, L, and T are finite. 2. An adjacency
relation v is defined on the set L, satisfying
the following properties: a) if lulj, then ljul;
b) for any position lx, except two (/, and h),

SAO there are two adjacent positions, i.e., there
- e exist two positions /, and /,(l,#/,) such that
Luly, Lvl, (le#ly, Isl; c) lp and lr have
use Ner $7 one adjacent position each; d) any partition
of L into two parts generates adjacent
positions which belong to different parts.

A message is probably always expressed
by some text. Cases when not all the letters
of the alphabet occur in the text may give rise
to some doubts. Anyhow, the concepts of "message" and "text" are largely
interchangeable, and we will assume that they are synonymous.

We will now proceed with the problem of identifying what messages are
worth decoding.

FIGURE 51. Example of an unconnected
text.

Artificial and natural messages

To distinguish the signals from "ordinary" stars and signals transmitted
by intelligent beings, we speak of "natural" and "artificial" messages,
respectively. It is often assumed that artificial signals from outer space
Should markedly differ from natural signals in some unusual property, which
cannot be accounted for by physical considerations (see Chapters I and III).

However, many quite unexpected phenomena eventually find "natural"
explanations; yet there are examples of artificial communications which can
be made as close as desired to natural messages.

Consider the hypothetical case of a high-quality 3D cinema. If the screen
is inaccessible, there is absolutely no way to distinguish between the view
through a window (natural message) and the view projected on the screen.
Note that the invention of holography will probably lead to development of
three-dimensional movies with precisely these properties.

Another example is less tangible, but it has bearing on the case of signal
Search in outer space: consider a variable star observable at point A and not
observable at point B. At the same time, point A is within the visibility
range from point B. An observer at À may inform an observer at B of the
exact behavior of the star by constructing a model which exactly simulates
the behavior of the variable star. If the quality of the model is sufficiently
high, the signals from the model will be indistinguishable from signals
emitted by the real star. Nevertheless, the signals from the model are
artificial, and the signals from the star are real.

° A graph is a union of two sets: the set of "vertices" and the set of "sides," in two-to-one correspondence
(i.e., each side joins two vertices. Graphs can be presented in graphical form: a typical example is an
airline flight network, with towns acting as "vertices" and flights as "sides."

136

IV. MESSAGE DECODING

These examples illustrate the futility of all attempts to devise a general
formal definition of the concept of artificiality to be applied as a general
criterion of signal selection.

We will now try to show that even legitimate artificial "meaningful"
messages may have a form which will rule out all possibility of decoding.

If there are signals of this kind, they will remain unintelligible despite their
probable artificiality.

Any message can be "scrambled" in such a way that it will be under-
Standable only to an observer with adequate "descrambling" knowledge, or
in other words an observer who has in his possession the "key" to the cipher.
In some cases, messages can be descrambled even if the key is not available
to start with. However, if the key volume is comparable with the volume of
the coded message, the text can be so scrambled as to become theoretically
undecipherable by any conceivable technique. This observation is due to
Shannon /8/. Examples of such scrambling techniques are easily constructed.

Consider a Russian-language text N letters long. The position of each
letter in the text can be specified by its running number ; from the beginning
of the text. Each number i (1 <i<N) is written on a separate card and the
cards are then shuffled and spread in a random sequence. In the resulting
sequence C, the card i will occupy position j from the beginning. If the
appropriate letters of the original message are substituted for these positions,
we obtain a coded message. To decipher the message, we require the
Sequence C (the key). In deciphering, the j-th letter of the coded message
Should be moved to a position identified by the j-th element of the key.

If the key is not known, this message clearly cannot be decoded; the
coded sequence of letters is truly a random sequence. Although the relative
frequencies of the individual letters correspond to the frequencies of the
Russian language, this fact can be easily concealed by adding as many rare
letters to the text as is needed to equalize all the frequencies.

Intelligibility of a message

We are thus concerned not just with messages sent by intelligent beings,
but with intelligible messages, i.e., messages that can be understood.

Are there specific criteria distinguishing intelligible messages from
unintelligible ones?

Suppose that only part of the text is available for examination, i.e., the
text has been partitioned into an accessible and an inaccessible part. If, by
examining the accessible part of the message, we can predict what the
inaccessible part probably contains, we will say that the message is intelli-
gible relative to the given partition. If the inaccessible part can be predicted
for any partition of the text into an accessible and an inaccessible part, we
Say that the message is completely intelligible.*

We will show in a few examples that this definition of intelligibility does
not contradict the usual sense of this word.

Indeed, the sentence "Pushkin was borr in the 18th paragraph" is unintel-
ligible because if the accessible part of the message is "Pushkin was born in
the 18th..." it is impossible to predict that the next part of the text is

* Intelligibility relative to a particular partition is a numerical function of the partition. No formal expression
for thís function can be given at this stage, however.

137

EXTRATERRESTRIAL CIVILIZATIONS

"... paragraph." One would naturally expect a word (or a group of words)
signifying a period of time (e.g., '... century"),

The sentence "Pushkin was born in the 18th centuries" is equally unintel-
ligible, since we expect the word "century" in singular and not in plural.
The sentence "Pushkin was born in the 18th siécle" is again unintelligible,
Since there is no reason to expect a French word in an English sentence.

(If the sentence is unintelligible, but it is clear how it should be modified to
make it intelligible, we generally say that the sentence is incorrect.)

A picture ofa man with the left leg replacing the right arm is unintelligible;
an object which looks like a log but sinks in water behaves unintelligibly; a
random sequence of letters is completely unintelligible.

Let us now consider examples of intelligible messages. An infinite
sequence of letters "... aaa..." is intelligible relative to any partition,
since the only reasonable prediction is "the i-th position of the unexamined
part of the sequence is occupied by the letter a," and this prediction is
always true.

A message of the form "... abcabcabc...'' is intelligible relative to any
partition for which the accessible part is long enough to reveal the three-
letter cycle. The picture of an infinite straight line is completely intelligible.
The sentence "Pushkin was born in the 18th century" is intelligible to an
educated English-speaking person, i.e., a person capable of predicting the
sequence of occurrence of words in English-language sentences, The picture
ofa manis intelligible to all intelligent beings who have seen a man alive or in
other pictures. Any periodic process is intelligible relative to partitions
revealing a sufficiently long part of the message.

We will now show that the ability to predict is based on knowledge of
certain special properties of the text or its components. Consider the
sequence of words "Napoleon invaded Russia in..." It can be completed
to read "Napoleon invaded Russia in 1812," but an equally intelligible
sentence will be "Napoleon invaded Russia in the 19th century." Formally
and morphologically the words "... 1812" and "... the 19th century" are
as far apart as the expressions "18th century" and "18th paragraph" in the
previous example. And yet there is a conceptual similarity between these
expressions, i.e., they fall in the class of words which are "close in
meaning' or, to use a different phrase, their "semantic distance" is small.

Texts may comprise small elements (e. g., words), as well as large
elements (e. ge. sentences). If we know what typical word combinations
make a sentence, we can predict the missing words having read through a
part of a sentence (typical examples are combinations of so-called gramma-
tical classes, e.g., the "nominative case," "finite verb," etc.). Correct
prediction thus requires breaking the text into sentence-like parts.

This partition may be quite complex; compound sentences are a common
occurrence in modern languages. We should therefore try to assess the
closeness of "words" not in terms of their "adjacency," but by some other

method.

138

IV. MESSAGE DECODING

Meaningfulness of a message,
predictive system, language

The information required for effective prediction of textual elements canbe
indicated by an appropriate re-coding of the message, whereby semantically
close parts are written in one common form, and the semantically dissimilar
parts are written in different form; textual elements combined into larger
components should be enclosed in brackets; "semantically close" parts
should also be textually close. This re-coding and rearrangement of the
text willbe called interpretation.

The best interpretation is clearly that which ensures the highest intelligi-
bility. The selection of the best interpretation may be regarded as message
decoding in the narrow sense of the word or as partial decoding. The
correspondence (mapping) between the elements of the message and the best
interpretation will be called the "predictive system" of the message or its
complete grammar. The predictive system, on the one hand, is close to
conventional grammars and, on the other, to dictionaries.

The language is naturally defined as the set of messages with the
same predictive systems. In other words, the messages in one language
are constructed "in the same way." If there is a correspondence between
the elements of the best interpretations of two messages, a certain corre-
Spondence also can be established between the elements of the messages.

In this case, one text is à translation of the other.

The translation of a coded message can be regarded as the ultimate aim
of decoding. We will see in $8 that it is easier to look for correspondence
between the elements of the messages than for correspondence
between the elements of the best interpretations. For decoding purposes,
we should therefore study the predictive systems of known, as wellas
unknown, languages.

The above examples of intelligible messages are disappointing to a
certain degree. Intelligibility clearly does not exhaust all the properties
of messages which have bearing on successful decoding. We will try to
make use of the fact that interstellar messages are probably constructed
in a Special way so as to facilitate decoding to a maximum degree.

The best interpretation in this case should be easily identifiable,
i.e., it should be readily distinguishable from the other interpretations. If
the quality of an interpretation is assessed in terms of its intelligibility, the
identifiability of the best interpretation can be defined as the difference
between the intelligibility of the best interpretation and some other (e.g.,
worst) interpretation. The identifiability of the best interpretation is a
fundamental property of messages intended for decoding; it is this property
that we call meaningfulness.

It is readily seen that messages without sufficiently intelligible interpre-
tations are not very meaningful; on the other hand, messages for which all
the interpretations are intelligible are not very meaningful either. This
accounts for the triviality of the examples described on p. 138: no low-
intelligibility interpretations can be constructed from these examples. Note
that messages expressed in normal languages (without any coding) are
"intended for deciphering" in a certain sense, and are therefore highly
meaningful.

139

EXTRATERRESTRIAL CIVILIZATIONS

Let us briefly consider the concept of "external meaningfulness."
Consider two partitions R; and R; ofa text T into an accessible and an

inaccessible parts, R:= T4", TPS; Rj- Tj^, T, Ti* c Tj^ (the symbol

Tj identifies the accessible part of the message, the symbol 7;"* the
inaccessible part). To each of these partitions corresponds a certain value
of the intelligibility II in the best and the worst interpretations, II(R;)*e* ,
II(R;)"***, II(R;)*** , H(R;))"*, The increment Ap of meaningfulness on passing
from partition R, to R; is expressed in the form

Ao = [I (R) best — II (Ry PA [I (Res —II (RJ).

The value of Ap can be defined as the external meaningfulness of a
message whose text is T7“\ T;^ (i.e., accessible in A; and inaccessible in
R; ). In particular, if T is the text of the message about the outside world
provided by the sensory organs during the entire span of human life, Ap is
the meaningfulness increment acquired as a result of a message with the
text T;/^N Ti^.

A particular example of the application of these principles will be
described in $6.

$3. TRADITIONAL METHODS OF MILITARY
AND LINGUISTIC DECIPHERING

Military deciphering

Deciphering of coded messages is a common practice in two fields of
human activity: it is often the task of historians and linguists (in their
attempts to read texts in lost languages), and also of military and diplomatic
personnel, who have to deal with intentionally coded messages in known
living languages.

According to the literature (see, e.g., /14/), military deciphering
techniques assume certain limited traditional forms (although, as we have
Seen, messages can be scrambled beyond all ability to decipher them).

A military cipher is difficult, and sometimes even impossible, to break.
These ciphers, however, are fundamentally simple compared to the predicate
system (grammar) of a reallanguage. Coding is generally done through
juggling with letter sequences which do not have any semantic relation to the
actual text.

Let us consider some of the common ciphers /14/.

Simple substitution cipher. Each letter is replaced with an
alternative symbol (generally another letter).

Transposition with a fixed period ¢. The entire message is
divided into segments £ letters long, and the same substitution is applied to
each segment.

The Vigenére cipher and its modifications. The key isa
sequence of ! letters. It is written consecutively, as many times as is

140

IV. MESSAGE DECODING

needed, under the original message, and the two sequences are added
modulo n, where n is the number of letters in the message alphabet.*

For example:

original message — LETTE RNOTYETRECEIVED
key — TROYTROYT ROY TROY TROY
cipher — EVHRXIBMMPSRKVQCBSMR

If the key is a single letter, the result is known as the Caesar cipher;
coding can also be done using an aperiodic letter sequence (which produces
an indecipherable cipher). In another cipher, each letter is replaced with
a sequence of ¢ symbols. In so-called "code systems," words, groups of
words, or syllables are replaced with various letter combinations.

Deciphering is based on two fundamentally different approaches: the
statistical method and the method of characteristic words. In the statistical
method, the frequencies of the letters in the cipher are compared with the
frequencies of the letters in the real language in which the message is
presumably written (the real language statistics is obtained from a suffi-
ciently large representative sample). If the frequencies of the letters in the
language are close to the frequencies of some cipher elements, these
elements are interpreted as the images of the corresponding letters.

In the method of characteristic words, we search for smaller component
elements which repeat like the letters of certain characteristic words which
are presumably contained in the cipher. These principles are used in the
algorithms of $8.

In certain cases, allthe possible ciphers of a certain class can be
examined, and the text being analyzed can be applied to verify that a
particular cipher has been used in that case. For example, if it is known
that the Caesar cipher has been used, the probability of a particular cipher
is a function of the intercepted cryptogram volume, and this function can be
calculated.

Suppose that we have intercepted a cryptogram containing the part of a
sentence "... creases to...'' (where "creases" is the end of the word
increases"), coded in the Caesar cipher. If only the cipher of a single
letter "c" has been received, the deciphered result may be represented by
any letter of the English alphabet. In this case, the probability of each
deciphering is equal to the probability of the corresponding letter.

If two letters (cr) are received, there are 26 different decipherings of
the message (assuming that the Caesar cipher has been used). The probabi-
lity of each version is equal to the probability of the corresponding pair of
letters in the English language, and so on.

Table 4.1 (due to Shannon) lists the results of these calculations for
sequences of up to five letters. Suppose that the enciphering was done by
using the letter a;. The letter sequences under the heading "Deciphering"

* Addition ofletters modulo a is carried out as follows: let i(a), i(b), and i(c) be the current numbers of the
respective letters in the alphabet. Then c is obtained from the equality i(a) 4-i(b) =i(c)if i(a) +i(b)S n,
and i(a)+i(£) — n-i(c) if i(a) +i(b) 5n.

The cipher is deciphered with the aid of subtraction modulo n, i.e., i(a)is found as follows: i(a) =
—i(c) — i(b) if i(c) — i(5) 20, and i(a) 2i(c) —i(b)-Fn if i(c) —i(b) «O.

141

EXTRATERRESTRIAL CIVILIZATIONS

TABLE 4.1
Deciphering Nel Na? N=3 N=4 N=5
CREAS 0.028 0.0377 | 0.1111 | 0.3673 1.0000
DSFBT 0.038 0.0314
ETGCU 0.131 0.0881
FUNDV 0.029 0.0189
GVIEW 0.020
HWJFX 0.053 0.0063
IXKGY 0.063 0.0126
JYLHZ 0.001
KZMIA 0.004
LANJB 0.034 0.1321 0.2500
MBOKC 0.025 0.0222
NCPLD 0.071 0.1195
ODQME 0.080 0.0377
PERNF 0.020 0.0818 | 0.4389 | 0.6327
QFSOG 0.001
RGTPH 0.068 0.0126
SHUQI 0.061 0.0881 | 0.0056
TIVRJ 0.105 0.2830 | 0.1667
UJWSK 0.025
VKXTL 0.009
WLYUM 0.015 0.0056
XMZVN 0.002
YNAWO 0.020
ZOBXP 0.001
APCYQ 0.082 0.0503
BQDZR 0.014
H (decimal 1.2425 | 0.9686 | 0.6034 | 0.2850 0
units)

are the sequences obtained by subtracting from the cryptogram the
sequences

a, âi, Qj, aj.
Qi- Qi- Qi- Qi- Qi ses
a, a, a, ai, a, iiy

Qs Qn, Gn, Gn, Any veers
Gist, Qis tet, Ansty Asst ++

The first column of numbers (N= 1) gives the probability of single-letter
sequences in the English language. These sequences provide the deciphering
probability of a single letter "c". The second column contains the decipher-
ing probability of the first two letters (cr), i.e., cr, ds,et, etc. The column
N= 5 contains the deciphering probability of all the five letters of the text
creas, i.e., creas, dsfbt, etc. In this column, the probability of the sequence
creas is close to 1, and the other probabilities are close to zero.

Vacant spaces in the table correspond to very low probabilities, The
probabilities were calculated from data about the frequencies of two- and

three-letter sequences given in /8/.
The row H gives the entropy of the probability distributions for the five

cases. The entropy H (H — — $p; log p;]was calculated from the values of p in
i

this table, using decimal logarithms.

142

IV. MESSAGE DECODING

Linguistic deciphering

The deciphering of old texts is apparently more relevant for the purposes
of interstellar linguistics. Students of old languages are forced to recon-
struct their highly complex structures. Moreover, the texts are not
Scrambled intentionally, and they are therefore far from a random jumble
of letters. The complexity of the natural languages, however, is responsible
for the lack of a general deciphering method, despite the successfulcracking
of numerous old texts.

The various cases of successful deciphering of old texts are largely due
to pure luck and to ingenious intuitive guesses, which will not work in other
cases.

Thus, the world-famous deciphering of the Egyptian hieroglyphs is
traceable to the discovery of bilingual inscriptions, i.e., the unreadable text
was accompanied by its translation; the Hittite language was deciphered after
a brilliant guess as tothe nature ofthe related languages; the Creto-Mycenaean
inscriptions were deciphered on the assumption (since proved correct) that
the language in question was Greek.

We will quote here from the article by Hrozny (who deciphered the Hittite
cuneiforms), describing the first breakthrough in his work. Note that the
pronunciation of the individual cuneiforms was known at that time, and the
meaning of ideograms — i.e., symbols representing concepts, and not
sounds — was also familiar.

"The method of my work is best illustrated by considering the following
Sentence, one of the first whose meaning I was able to establish, and in
which I recognized three Hittite words of Indo-European origin.* This
cuneiform I read phonetically. **

ru XV -an ezateni vādar-ma eruteni.

"When I first came across this Hittite sentence, I knew only the meaning
of the ideogram, which often, though not always, stands for "bread." Other
parallels indicated that the suffix was accusative singular. Despite numerous
other possibilities, it was reasonable to assume that a sentence dealing with
bread will also contain the verb "to eat." Itherefore started with the purely
hypothetical assumption that the word "ezateni" signifies the concept of
eating. Soon after that I noticed that the Hittite root "eza" stands for "to eat"
in many other texts, and that another root with the same meaning is "ad,"
e.g., in the form "adanzi," they eat, which is probably identical with "eza."
Then I compared, again purely hypothetically, these Hittite roots "ad," "ez"
to the Latin "edo," the German "essen," etc. Other sources supplied me
with an indication that "teni" is a second person plural ending in present and

* Related languages are languages arising from a common "source language." Words of close meaning in
related languages have a similar pronunciation.
Russian, Ukrainian, Polish, Czech, Bulgarian, and Serbian are closely related languages (the so-called
Slavic languages); more distant relations of Russian are German, Latin, Greek, and some Indian languages,
forming together the so-called Indo-European family. Compare the following words:

Russian MaTb

German Mutter
Latin mater
Greek pnto
Sanskrit mátar.

** Le., its pronunciation was known. The «uneiform in the middle is an ideogram (a concept symbol). The
phonetic composition of the equivalent word was not known.

143

EXTRATERRESTRIAL CIVILIZATIONS

future tense, so that I translated the first sentence as "you will eat bread."
The next sentence looked parallel to the first: "vadar," a noun; "ma," a
preposition; "ekuteni," a verb with "teni" ending. Since the word "vadar"
was parallel to the word "bread," it probably also identified some simple
food. The English word "water" and the Anglo-Saxon "watar" helped me to
identify "vadar' as water.

"The noun "water" was thus followed by the verb "ekuteni," which corre-
lated with the verb "ezateni," "you will eat." It therefore logically lended
itself to translation as "you will drink." Later I found that besides the root
"eku," to drink, there was also a close root "aku," to drink, e.g., in the
word "akuvanna," to drink, The comparison of "akuvanna," to drink, with
the Latin "aqua," water, was self-evident. I therefore translated the whole
sentence as "you will eat bread and you will drink water."

It is clear from this excerpt that there could be no continuation to
Hrozny's method: his guiding line was the assumption that the lost Hittite
language was related to some known languages (words of similar meaning
have close pronunciation in related languages); the rest of his arguments
are fairly obscure, e.g., the contention that the two parts of the sentence
are parallel and the frequent references to similar meaning that the same
word has in other texts.

If the pronunciation of the letters is not known either, we have to rely on
the occurrence of proper names in the text, as they are of international
meaning to a certain extent; ideograms and hieroglyphs (whose meaning in
a sense corresponds to that of a picture), pictures occurring in the text, or
objects carrying inscriptions are very helpful in disclosing hidden meaning.
There is generally some information available about the corresponding
historical epoch, the wars which took place at the time of writing of the texts,
the identity of rulers and leaders. Whole dictionaries are sometimes avail-
able (the Tangut and Mayan inscriptions). The decoding of old texts is thus
largely dependent on the resourcefulness and the intuition of the linguist, who
draws upon a tremendous treasure of information that may prove useful; all
this, however, does not provide us with a set of general linguistic tools and
techniques for text deciphering. Deciphering is closer to a one-time art,
not quite understandable to the outsider, than to a practical science.

$4. SEQUENCE OF APPLICATION AND STRUCTURE
OF DECODING ALGORITHMS

Sequence of algorithm application. Levels

In decoding a particular extraterrestrial message, we shall naturally
have to lean heavily on our intuition and more or less incidental information.
However, insofaras no extraterrestrial messages have been received, there
is only one way for us to prepare for the future decoding, and this is by
developing general decoding methods which will answer the greatest variety
of needs.

In our opinion, these methods will be valuable only if they admit of clear-
cut, unambiguous formulation. This condition is met by algorithms —

IV. MESSAGE DECODING

precise instructions for textual analysis which are so clear and comprehen-
sible that a computer can carry them out. The computer-oriented approach
is particularly helpful since deciphering involves processing of large blocks
of information, which often cannot be done manually.

Given a complete system of such algorithms, we can visualize the
operation as follows: the text to be decoded is fed into the computer, which
then proceeds to translate it into one of the known languages. This ideal
situation, however, is not very realistic.

There is, moreover, no need to carry the algorithm system to this
extreme: it is sufficient to ensure algorithm solution of some "key"
problems of decoding. This will leave relatively simple problems to be
tackled by human ingenuity and intuition, the two properties presently
unattainable by computers.

As we have noted before, decoding is primarily an activity intended to
identify the "predictive system." Its second aim is to translate the original
into one of the known languages. If there had been powerful decoding
techniques meeting the second aim, we would not have had to search for the
"predictive system" of the original.

Practical experience shows, however, that it is nevertheless better to
Search first for the predictive system and then to proceed with the develop-
ment of translation techniques. Experiments carried out using the algorithm
on p. 179 are highly illustrative in this respect. Moreover, the possible
"non-interpretability'" of extraterrestrial messages should be taken into
consideration (see conclusion).

We will consider some algorithms whose importance for the construction
of the predictive system is self-evident.

Traditional linguistics uses two techniques to distinguish between linguis-
tic phenomena: one of these techniques resorts to real images and patterns
that various expressions invoke in the mind of the language user, and the
other makes use of our inherent ability to differentiate between correct and
incorrect expressions in a particular language.

For example, verbs are distinguishable from nouns because verbs
generally express a certain action or process, whereas nouns are identified
with objects or abstract concepts. On the other hand, morphologically, a
[Russian] verb is identifiable by its characteristic "endings," such as "a",
naan, tao (endings of past tense masculine, feminine, and neutral).

Modern applied linguistics, with machine translation as one of its most
active branches, uses mainly information belonging to the second category,
i.e., advance knowledge of certain morphological signs of linguistic
phenomena is presupposed. .

In the decoding of extraterrestrial messages, we naturally cannot resort
to real images or patterns or to morphological features of the written
language.

In the construction of decoding algorithms, we should proceed from the
basic and most general properties of the phenomena. It is here that the
linguist's interests lie.

An efficient decoding algorithm essentially provides a definition of the
phenomenon that it is supposed to recognize. More precisely, we could
simply define a particular linguistic phenomenon by what emerges from an
arbitrary text when a particular decoding algorithm is applied to it.

These definitions are attractive in that they are applicable to unknown
languages (i.e., they are highly general), they are extremely lucid (and can

145

EXTRATERRESTRIAL CIVILIZATIONS

be implemented by a computer), and are practicable, i.e., they provide a
tool for recognizing various linguistic effects. The importance of these
algorithms may turn out to be quite independent of the linguistic decoding aims.

Let us consider in more detail the structure and the sequence of applica-
tion of decoding algorithms. Decoding algorithms clearly may use informa-
tion disclosed by other decoding algorithms. If a certain algorithm B uses
the ability to recognize a linguistic phenomenon defined by algorithm A, we
will say that algorithm B is of a higher level than algorithm A. It would
naturally be very unfortunate if algorithm A were at the same time of a
higher level than algorithm B, since this would lead to a definition of an
unknown in terms of another unknown. The only exception are algorithms
which successively improve their own results. In this case, the seniority
of the algorithms is determined by their seniority in the first iteration.

It is clear that there must exist a zero-level algorithm which does not
use any information obtained by other decoding algorithms. Zero algorithms
Should differ according to the effect that the symbols have on the human
sensory organs or on the decoding device. They should be associated with
the minimum differences detectable by these sensory organs.

If we are developing algorithms intended for the analysis of written
languages, the zero algorithms should naturally reconstruct the alphabet of
the particular language by examining a certain sufficiently long text. The
information required apparently reduces to the ability to distinguish between
black and white squares, assuming that the text is covered with a very fine
grille so that each cell is either black or white. The ability to identify the
position of each cell is also required.

If spoken sounds are to be decoded, the zero algorithm should use the
minimum acoustic differences. The variety of the zero algorithms evidently
can be reduced by suitable conversion of signals with physical devices; e.g.,
Speech can be represented by a chart plotting on paper the variation of air
pressure. .

At first glance, alphabet reconstruction is a very simple problem, which
always can be solved after a brief inspection of the text. The human analyst
is sometimes baffled by illegibility of the written text, but for machines the
problem is complicated even for fairly clear inscriptions. Some insight into
the problems involved in the identification of the phonetic alphabet may be
gained by inspecting a segment of an oscillogram trace of Russian speech
(Figure 52).

A curve representing a signal from outer space will be much less "legible":
it will probably be distorted by strong noise. Curve manipulation is not
among our strongest aptitudes, and it is therefore clear that the reconstruc-
tion of the alphabet of "elementary" signals will not be an easy undertaking.

The zero algorithm for written languages is thus expected to reconstruct
letters as special combinations of dark and light squares. Although efficient
algorithms for alphabet reconstruction can be developed in principle, no
such algorithm is available at this stage. In $11 we will describe a rudi-
mentary algorithm of this kind which is more of theoretical than practical
importance.

Once the set of "elementary signals" has been identified, we can proceed
with identification and analysis of larger elements. For languages close to
human languages, the first level algorithms should be able to distinguish
between various classes of letters of similar pronunciation, and also smallest

146

IV. MESSAGE DECODING

o r

TI Ld L3

DH HHH! ina áj

FIGURE 52, Oscillograms of Russian spoken syllables "tu," "ta," "pa":

All the syllables are stressed, extracted from a recording of individual sentences. The
flat portion of the oscilloyrams corresponds to silence (closed mouth), then follows a
burst which cnsures the audibility of the sounds "t" and "p." and further large-amplitude
fluctuations representing the vowels "u" and “a.”

meaningful letter sequences which are not made up of smaller meaningful
sequences (the so-called morphemes). This algorithm should be able to
divide the message into morphemes even if no blanks are interposed between
the words, since words are more complex elements than morphemes. In
some orthographies, no division is made between words anyhow.

Second-level algorithms should locate the limits of the individual words
and identify different classes of morphemes (such as semantically meaningful
morphemes and auxiliary morphemes used as suffixes, prefixes, etc.). The
third-level algorithms should search for classes of words and identify the
limits of sentences. Higher-level algorithms should analyze those sentences
and semantics.

Numerous algorithms of different levels may be quite similar. In some
cases they are actually identical, differing only in the input material. Thus,
algorithms identifying groups of letters with similar pronunciation can be
used without any modification to identify different classes of morphemes;
sentence-identifying algorithms are very similar to syllabilization algo-
rithms; algorithms splitting the text into morphemes are not unlike the
letter-identifying algorithms, etc.

147

EXTRATERRESTRIAL CIVILIZATIONS

In our discussion of the particular algorithms, we will always indicate
on what different levels the particular algorithm may be used. However,
essentially similar algorithms may each have its own specific features,
generally related to the volume of processed information.

For example, an algorithm identifying classes of morphemes willprovide
an output which is about a hundred times larger than the output of the same
algorithm operating in the letter-identifying mode. Because of these specific
features, the programming of higher-level algorithms is substantially more
complicated.

On the other hand, higher-level algorithms are naturally more interesting:
they provide a fuller analysis of the text, permitting "long-range forecasting."
We will consider low-level algorithms, e.g., algorithms analyzing letter
pronunciation. Not all of them are relevant for the decoding of extraterres-

trialmessages. However, they should all be considered as models: on a
low level these algorithms solve problems which are much more topical and
significant when tackled on a higher level.

There probably exists a limited number of different types of decoding
algorithms, and we should therefore first examine one algorithm of each of
the different types, before striving toward higher and higher levels.

Structure of algorithms: sets of alternatives, quality
function, computation procedures. Types of algorithms

Various decoding algorithms have many features in common. We have
indicated earlier that algorithms recognizing different linguistic phenomena
in an unknown text may be used to provide the definition of the corresponding
phenomena. However, somewhat more general definitions also can be
offered. To this end, it suffices to formulate clearly the characteristic
features of the linguistic phenomenon used in its identification. Computation
procedure intended for recognition purposes (e.g., algorithms) may take
different forms even for the same set of recognizable features. In our
description of the decoding algorithms, we shall first describe the recogniz-
able distinctive features, and then give the particular recognition procedure.

Recognition features in their turn fall into two categories: some do not
require any computations or manipulations of the text, whereas others do.
The former features are of binary character, i.e., they are present
for a certain phenomenon and absent for other phenomena. Features of the
Second category express properties which are more prominent in this partic-
ular phenomenon than in other phenomena with the same "binary" recognition
properties.

We say that features of the first group define the set of alternatives (the
set of interpretations), whereas the features of the second group charac-
terize the quality or the reliability of these alternatives. In other words,
quality is a numerical function defined on the set of alternatives. The set
of alternatives will also be called the set of permissible solutions, with a
certain quality function.

We wish to emphasize one highly important property of quality functions.
Until recently, linguists used definitions based on binary features (or, in
general, features expressible by a finite number of digits). These definitions.
however, proved to be quite complex: they contained numerous "exceptions"
and were not particularly suitable for machine recognition.

148

IV. MESSAGE DECODING

This approach precluded the formulation of common definitions for
similar phenomena in different languages. The concept of quality function
greatly simplifies the "binary," i.e., logical, part of the definitions, and
they acquire a greater generality. The reader will see that quality functions
proved highly convenient in practice, since algorithms using these functions
are generally programmed without much difficulty.

The aim of the recognition procedure (the algorithm) is to find a permis-
sible solution which maximizes (sometimes minimizes) the quality function.

Whenever the set of permissible solutions is given and the quality function
is defined, the determination of the permissible solution maximizing or
minimizing the quality function becomes a purely mathematical problem.

Rigorous solution of mathematical problems of this kind to which decoding
algorithms are reduced is mostly unknown. We tried to describe the most
practical solutions, i.e., solutions which are sufficiently accurate to provide
acceptable results and yet sufficiently simple to be implemented on existing
computers.

Let us again consider the question of the various types of algorithms.

The currently known algorithms intended for the analysis of predicate
systems can be divided into the following groups:

Classification algorithms. These include the algorithms which divide the
set of units being studied into nonintersecting subsets, e.g., the algorithms
partitioning the set of words into classes which contain letters of similar
pronunciation; algorithms partitioning the set of morphemes into classes of
morphemes with identical "grammatical" properties (auxiliary morphemes
vs. meaningful morphemes); algorithms identifying semantically close
classes of words.

Matching algorithms. We use this term for algorithms which form small
linguistic elements into larger linguistic units; e.g., the algorithm of mor-
pheme identification, the algorithm of letter identification, the algorithm of
Sentence identification, and the algorithm of syllable identification.

Algorithms establishing semantic closeness. The visual closeness of
words in a text does not always correspond to the actual semantic closeness
of words. Similarly, in a linear scan of a two-dimensional pattern, adjacent
elements are not the only close elements: elements separated by the length
of one line are of course also close.

Algorithms of this kind include the algorithms which determine the so-
called sentence graph (see p.198). Note that knowledge of the "true close-
ness" of elements is essential for correct functioning of the matching
algorithms.

Translation algorithms receive less attention in this chapter. Decoding
apparently can be confined to algorithms compiling various bilingual
dictionaries. In machine translation, algorithms synthesizing sentences in
the product language are of considerable importance. In decoding, this
problem can readily be left to the human operator.

The description of the various algorithms in this chapter does not corre-
Spond to the order indicated above. Simpler and more obvious algorithms,
accompanied by examples, are given in $5 and $6, the others are deferred to
§7 through $11,

149

EXTRATERRESTRIAL CIVILIZATIONS

eee

0000*—t-*t—*t—-*t—00000000000000000000000000000 song
*-*-000000000000*-000*4-—*t—-tt-—tk—--04-44—-—D0 pozi[e1e|eg
000000000000000*—-—-*—-0-*-*4—-00-*-*-*—4-0004-—4-0 1ue1ouos
**t-00000000000-4———--*—00000000004444—0000000 snonuriuo:)
0000000000000000000000000-*4*——--00000*4----Q0 lesen
000000000000000000000-*4*tt------ Td b44dkb-c--- 0 3502,
0000*tt——00-*4*t—-—^4t4t—-——-tLx 4 t d vl dltlnfóAàtktie SER qoe ce v e ERIT 0 yard [oA2]- ^0]
00000000+r----+t+t++t+t++--------------------- 0 12eduio?)
0000-ct4t-—----- 00000000000000000000000000000 uodo
FPE PE He s eri eH eK Tob GB GB G BR BR VB B 4: 4-4: 4 ttt ttàtàtdcbt4à4e6Akà-— Qnuruosuo
bh ht eb b bb b bg ge—-ee-——el2222-ce-2c22222l222222 91T€20A
4 db a3 gn d ,ee?,o0, 9 x Hx» 2 s 2,5, A J| M w,qd,d q dz,sz so»,uu,papifl

wr d d anon gon ee 0 O0 2 2 » V 3 3 X m » ,8,p u$ KW M ,g,dg d fot o» w aum were y

eee

o'y ITAVL

150

IV. MESSAGE DECODING

$5. CLASSIFICATION ALGORITHMS (PART I)

Distinctive features and classifications

Classification algorithms permit the assessment of similarity and dissimila-
rity oflinguistic phenomena. A linguistic unit is generally characterized by a
certain Selection of properties or distinctive features which are present in
the particular phenomenon and are absent in others.

If these properties are given, the particular linguistic phenomenon can
be described by assigning to it a vector of ones and zeros whose i-th coor-
dinate corresponds to the i-th feature; it is equal to 1 if the particular object
has the corresponding property and 0 otherwise.

It is sometimes assumed that the features may take on other values
besides 1 and 0. In general, a distinctive feature is a certain numerical
function defined on the set of the relevant objects.

If the feature may take on k values, it partitions the set of objects into at
most k nonintersecting classes. Conversely, if there is a classification
(partition) of the set of objects into & nonintersecting classes, one can
introduce a feature which takes on k values. This inverse line of reasoning
is characteristic of the decoding approach.

If two objects are described by the corresponding vectors, the similarity
of the objects can be estimated by calculating the distance* between them as
between points of n-dimensional space.

Table 4.2 /13/ characterizes the sounds of the Russian language.

The columns correspond to the phonetic letters of the Russian. A prime
next to a consonant expresses soft pronunciation, a prime next to a vowel
indicates that it is not stressed. The different features are listed inthe hori-
zontal rows. Vocality is assigned one of the two symbols + (vowel) or
- (consonant) for each letter; the letter "j" in the author's opinion is neither
a vowel nor a consonant, whereas "r," "r'," "1," "]'" are both vowels and
consonants at the same time. "Therefore, consonance is not specified by the
value of vocality. Stress is a feature characteristic of vowels only, and for
consonants it therefore takes on the value 0 (inapplicable).

The values + and - of a certain feature correspond to a greater similarity
of sounds than + and 0 or — and 0, and the distance between the sounds may
therefore be described by the function defined on p. 193.

If we want to apply the Euclidean distance p,,, = V(x, — xj), we should first
assign a certain number to each value of the different features (0, +, and —).
If we measure the distance between two sounds using equation (4.5) (p. )
the distance between a and b will be 20, and the distance between a and o only
3. 'This agrees with the intuitive concept of similarity of sounds.

If we have a selection of so-called grammatical classes of words (e. g.,
"nominative case," "masculine," "singular"), we can construct an analogous
table expressing the grammatical properties of words. Given a selection of
classes of words with some common semantic denominator (e.g., animation,
greatness, intelligence, etc.), we can construct a semantic description of
words.

* Distance is a function of pairs of elements of a certain set with the following properties: 1) p(a,a) =0
(nondegeneracy), 2) p(a, b) p (b, a) (symmetry), 3) p(a, b) +p(b, c) 2 p(a, c) (the triangle incquality).

151

EXTRATERRESTRIAL CIVILIZATIONS

Besides providing a convenient means of assessing the similarity of
objects, the description vectors can be used to replace the tremendous
variety of objects with sequences comprising a limited number of distinctive
features. Thus, using binary features, we can describe a set of n objects
with the aid of [logen]+1 features.* This is a highly valuable property of the
vector approach, seeing that the total number of various words and concepts
is really enormous.

Examining Table 4.2, we note that the features cover a wide spectrum of
properties: some of them are related to pronunciation, the others to acoustic
properties of sounds.

If we were to construct a similar table from an analysis of the various
combinations of sounds in fluent speech, we could reconstruct the sounds of
the various letters from written text. After all, written language does not
markedly distort the ability of sounds (as expressed by letters) to combine
with one another. If similar tables were available for individual words, in
Such a way that classes of words corresponding to a certain value of each
sign contained words with some common semantic denominator, we could
"guess" the meaning of words from an examination of texts.

This problem encounters considerable difficulties. Therefore, the
general scheme will help to better understand the classification algorithms
described below.

Algorithms for the identification of vowels and consonants

The first algorithms of this class reconstruct the pronunciation of letters
from the occurrence of their combinations in a text. It involves the partition
of letters into two classes using a single binary feature.

If this algorithm is applied to letters, it will identify vowels and conso-
nants; applying the algorithm to morphemes, we can distinguish between
meaningful morphemes and auxiliary morphemes. Application of the algo-
rithm to mathematical texts would differentiate between predicate symbols
(+, —, =, etc.) and object symbols (e.g., x, x, 10, 29?) When applied to
words, the algorithm will probably differentiate between nouns and verbs.

By identifying the vowels and the consonants one naturally does not
establish the exact pronunciation of the letters. However, this is a first
useful step toward decoding. **

Thus, if the algorithm is applied to letters, it provides a definition of
vowels and consonants. This definition is superior to conventional definitions
(of acoustic or physiological bias) in that it is applicable to letters for which
these traditional concepts are invalid.

The set of alternatives. The vowels and the consonants are thus regarded
as a certain partition of the set of letters into two classes: the class of
vowels and the class of consonants. In other words, it is assumed that these
two sets are disjoint and between themselves exhaust the entire alphabet.

This restriction, however, does not quite correspond to the true state of
things. Indeed, the letter "y' in English is sometimes rendered as a vowel
and sometimes as a consonant (compare "very" and year"). This is by no

* Here [loge nJ stand: for the whole part of the logarithm.
** The deciphering of inscriptions in the so-called Carian language carried out by V. V. Shevoroshkin began
with the identification of vowels and consonants.

152

IV. MESSAGE DECODING

means a result of some imperfection in the written language (in Czech also,
"p'" is a consonant in the word "Praha" and a vowel in the word "prst,"
finger). If we do not restrict the analysis to letters, we see that this is a
very common phenomenon; in particular, a single word often has a variety
of meanings (the phenomenon of homonymy).

However, it would be impossible to develop an algorithm for the identifi-
cation of vowels and consonants without these restrictions.

However, by stating that vowels and consonants constitute disjoint classes
of a certain partition we have said very little. If a particular alphabet
contains n letters, we may construct 2" different partitions! Nevertheless,
this statement is one step forward: so far the set of alternatives has not been
restricted at all.

The quality function. The quality function is constructed from the follow-
ing considerations: in any text, vowels are not very inclined to combine with
other vowels and consonants with other consonants. Conversely, vowels
readily combine with consonants.

If we take an arbitrary partition of the alphabet into two classes, we are
not likely to notice this property. Suppose that the letter P has been declared
as a consonant, and all the other letters of the alphabet as vowels. Under
this partition, "vowels" may clearly occur very often in close combinations.

Let us analyze the combinations of letters of some language using a table
whose rows and columns are identified by the letters of the corresponding
alphabet. The entry corresponding to the row i and the column j contains a
number which indicates how many times the letter a; and the letter a; occur-
red one next to the other in a given text (the order in which the two letters
occurred is immaterial).

Consider a certain partition of the alphabet into two classes. Allthe rows
and the columns headed by "vowels" are shifted to the left-hand top corner of
the table, which is separated from the other letters by a line. The table thus
takes the form

vowels consonants

vowels

ec -EN

Block 1 contains numbers which show how vowels combine with other
vowels, block 3 contains numbers which show how consonants combine with
consonants, and blocks 2 and 4 contain numbers showing how vowels combine
with consonants. If the partition is close to the true division into vowels and
consonants, the numbers in blocks 1 and 3 should be small, and those in
blocks 2 and 4 large. The quality of the partition therefore can be estimated
in terms of the sum of the numbers in blocks 1 and 3, say.

If the alphabet contains n letters, of which m are vowels, the correspond-
ing quality function can be expressed in the form

K= 2 à 9(a;, a)+ M, P3 @ (az, aj). (4.1)

153

EXTRATERRESTRIAL CIVILIZATIONS

Here (ai,aj) is the number of joint occurrences of the letters a; anda;,
regardless of order. The fact that we ignore the particular order in which
the two letters combine is indicated by the comma, thus (ai, aj) = @(aia;) +
+(a;a,), The letters a; and aj belong to one of the classes, and the letters
a, and aj to another.

The smaller the value of K,, the better is the partition. 'The best partition
is that when the function is minimized.

The above estimate function for the detection of vowels is not the only
possible one. "Various equivalent estimate functions are available, which
give an extremum for the same permissible solution which minimizes K,.
There are also interesting estimate functions which are not equivalent to K.
One of these is
m m
D Dp (ay, a)p*(a)4

1 j=

t
n

n
jl 2
a5, a) p^ (ay) —
k-n4l 2, P (as, a) p* (a)
m n

-22,XÀ plan ar) plas) p (aj. (4.2)

i=| kam+
n
> (ax)
The symbol p(a,) stands for Mo where a, belongs to the same

class as a, and a. This notation is based on the fact that the appearance of
any letter of a given class can be regarded as the appearance of some letter
> 9 (ay)
ax; Similarly, p(a,) stands for ar . The number p(a,) is the relative
frequency of one of the classes, and p(ax) the relative frequency of the other
class.
The function K, is similar to the function K;, which is equivalent to K,,

m m

K=% D p (ai, a)

isl je
D p(ea)-2X3 X plan a), (4.3)

kem+! l=m+ isl &em«l

differing from it in the coefficients p(a.) and p(a,). On the whole, the func-
tion K, is the correlation moment of the sequence of numbers 1 and -1
generated when 1 is substituted for each vowel in the text and —1 for each
consonant. The function K: reflects the nonuniform frequencies of the vowels
and the consonants. Combinations of consonants are more frequent, and
therefore less significant; this is reflected in the weighting of the occurrenc’
of consonants by the frequency of vowels, and vice versa. All this is highly
hypothetical, however, since experiments were performed with the function
K, only.

Recognition procedures. The simplest procedure based on the above
features is quite trivial, It suffices to construct a table of the frequencies
of pair combinations, examine allthe possible partitions, and evaluate the
estimate function for each partition. The partition corresponding to the
minimum value of the quality function is then chosen. However, the volume
of computations involved in this algorithm exceeds the ability of the largest
modern computers for alphabets of normal size (e.g., about 30 letters).

$780 154

IV. MESSAGE DECODING

The search for an effective procedure of minimizing the quality function
is associated with considerable mathematical difficulties. The choice there-
fore lies between impracticable and incorrect methods.

We will describe an algorithm which often minimizes the function K, fast
and without difficulty. For some tables it gives incorrect results, but even
these apparently are not too far from the best solution. Anyway, experiments
with this algorithm never led to errors which could be attributed to algorithm
imperfection. This imperfection emerged only when specially selectedtables
were used.

We describe the procedure step by step:

1. Foragiventext, construct the table of the numbers 9(a;, aj), where
9 (a, aj) is the number of occurrences of the pair of letters a; aj, irrespective
of order.

2. Cross out numbers of the form q(ai, ai).*

3. Compute the sum of numbers in each row of the table.

4. Move to the first position (left-hand top corner) the row and the column
with the largest sum.

5. Separate by vertical and horizontal lines the row and the column that
were moved from the other rows and columns.

6. For rows below the horizontal boundary, calculate the sum of numbers
lying to the right of the vertical boundary and the sum of numbers to the left

of the vertical boundary; subtract the second sum from the
first. The resulting numbers are called the decisive

CP 7 differences.
2 MA 7. If there are positive decisive differences, move to 9.
8. End. Rows above the horizontal boundary correspond
V és) to letters of the first class (generally vowels), and those
ED below the horizontal boundary correspond to letters of the

second class.

9. Select the row with the maximum positive decisive
difference and move it across the horizontal boundary; the
corresponding column is moved across the vertical
boundary. Return to 6.

Consider a small example illustrating the application of this algorithm.
Note that the algorithm can be used without separating successive
words. In particular, suppose that we do not know which letter is the first
letter of a word and which is the last letter (the word is inscribed along a
circle, as in Figure 53).

The table of the numbers g(qa;, aj) for this text has the form

FIGURE $3. The
word "parera",

dH mU A o» Uu

* These numbers are the frequencies of pairs of identical letters. They clearly enter the sum > > 9 (ai, aj) +
i l

a > > (ar, a1) for any partition of the alphabet into two classes, and therefore do not affect the quality of
k l

the particular partition.

155

EXTRATERRESTRIAL CIVILIZATIONS

After step 2, the table does not change. We proceed with step 3:

Piety sp cp es
A 2 | | 1 | [à 4
xy [i] ja
ej] did Jile
SNE C vif
Now we proceed with steps 4 and 5:

A P K E T
A 2] 1 | J 4
ee Eod
ER UN IN RUN
E NE Tod

Step 6:

A P K E T
aj f2] if o]:
Er pb j l ] -2
«rt p odi] fe
E WE m LL ps | 2
EEA FE e qo

From step 7 we move to step 9. Carrying out instructions 9 and 6, we get

A E P K T

From step 7 we move to 8 and end the analysis. The result shows that the
first class contains the letters A and E, and the second class the letters
P, K, T.

Table 4.3 illustrates the results of a similar machine experiment using
Russian, English, and French texts of 10,000 words each.

The results for the Russian and the French texts are virtually error-free.
Note that the Russian letters » and » correspond to vowels which have long

156

ODING

MESSAGE DEC

IV.

0000001] 0 0 I Zz 000000000 @
09010200 ! vr 0 6 0 cl S € 00 0 0 ZO
0101002; cl 2@ zZz age S8 € 00000 € 0
0010000 0 i 0 8 1! 0 O 110 80 v 0
060000€ 9 09 6 @ 6 S € 20 0 1 8 p
0000001 00 0 0 0 0 0 000 0 8 0 0
10 2%0¢€ Lr Of 8 8 Z 6 8 S £9 0 012 BZ
ot 60 9 0 Of SI OF ZI E IG eb ws ec GO 1 4 i P GS
ot & 1 020 8& 91 O 6S ZI cl v O R G OI % GI 1% 6
1020908 21 66 0 4 OF % 0l ago b 0| Z0
ZZ t18 6 0 0 z Z 2 08 & Ef U HU zi Ge ME
00€ LG OS I cl or € 9 OF 8 V HO S8 9 9 O
0 02180608 & V. 1% ef Ol @ 6l 9 Ol t 6 OF OI cl
0o¢e€0¢0 SI L OF OF Il 8 Gt b UO Gr € Le
0ogoretoel e 8 zl rl 9 HW O€ 0 &@ * S 6
00012809 2 S £ Z FI OF 9 £ 028 OS 06 !
0000000 1 oOo ZO èp ee 0 @ 0 & 0 I O
0008000 4 vé & 6 8 6 k 6G Orel 6 6r vo I
0000717 82 i|! 6 | 4 9 9 € & GO 60 S 0
0 ZE} 8 0 1% bb I 2 9) 9 OF 21S 021 WS 8 Z£
2$000F*02z S 6 0 € 0 @ 7 & 10 1 026 0
SS See a gee SORES he ae neo ee « "—— E
ro 10008 S ! b 3 | 6 6 O € O 2 0€ $ |
0000008 | 0 1 0 0 2 € 100 0 0 F 0-
0F0€g&£90 S 8 8 4 Sl ef € OL» O Tl I et V
11106 1 Ie o9 Il wj Ob Ol o€ OF | 9 b SIO Ik 8
0S I+ z2ost 4 tl 4 eb E eG V 19 0 8 t 2 9
J 2012 00 6 06 8 OF 6 cb O6 O F D Glo BS
8 O191% OLO £S PE Gb Le Eh 90 ví Ih O TI BI Gb 9% O6 06
13 6e € 81 gc 9 S6 68 89 Sh EvI ZOl SPl C8 Ie ze 9% 09 SZ G6 €6
¿Z ZZ €9 61 910 Vcr 6ZI 9ST IZ 20% 28 99 EF PE Gc bh £6 LE f8 OF
OII £01 OZI OF Scc 49 ELI 1v! OI 06 OF 66 tb ¿3I SE

6 09096 ab Il

9 9 I£ *. FhO ZOIL Cel IG] sT
BR ANE ELM MN COMES qe M MM MERE M AL ETUR
y 8&8

mmhn Xo

1 29 d U

18] SIl 261 €» 99 OE be

W mf Hu € kK

S1Uu?PUOSUO:)

401 961 791 c7

aeaee a e ÓMÀ

$100 10 L8 Io ZG 9 m
0 0 b ! S 2 96£ 2% OS $ m
100 ! I Ororep £9 09 IE h
oogcore2eza 6 6 + n'
0 09 * ZO Ol gc 9! c» v X
00010008 0 I 9 o
el £z bl Ig ST 6S9 eS s6 Fal OTT 261 L
G | S 094 OL be 68 660 EO! EC! 2
10 8 IL bl 6 cb 89 MST OZI IGT d o
y | 8 bre 8 22 IL 9% cel ues
Z 0 LE 0b £P OF Cb €b! 202 82% IST H 2
1 0 Storie 6 92 c01 28. 79. 81] w os
6 Z £I OF co cb Va BHI 99! ELT Z6I os
Z £ £ OL FP OZ IF C8 EF Zbl £I Yoo
0 | OLI 010 O le ve OI 9c u
£ 05 99 F IN Ze St 06 0€ €
000 P 0 0 zig Hr OP VC x
Z 0 licer 8 Bt 209 £6 66 ZO! Y
00 191Z% £ £ Æ J
£ b £I Ig 431203896 £8 Zot Z9 g
G 0 FIR 93S Oc cc 9% GE CA sg
S ENTIRE NOE Se hea
200009682 * & 4 OO,
00000€g£€&€z Ft € ,€;
000000! & Bt €& Id
00098 I! IL» ge S 08 9I K
000100226 set € q z
9€0100e6 S 2 V q 9?
66e t! * 2 € O0 & 2 OZ & oo
ZZ 902 6 LLG LS o6 GE Hoa
b b 61S SIS 4 2g 86 OD 6 8
€!à 1 OE T1 a Olea Ol 9 SI v
L£& 91€ F LÆ 66 SI 86 (0)
acaqKaqaaAndad vo ,
SJOMOA

ae Aa a M EE

uvissmy

'e'v ATAVL

157

ooo
ooo=--o

-~
e
N

OGONOTO SOD ONN ODO ONO OC ONC SC So
NOM eR ONDNK OHH Ds HB NO TOS
oo ooo oo co ooceoeaeoee Ss

oonoooo "m7 ONO SO
oono ~oonowrooc ©

EXTRATERRESTRIAL CIVILIZATIONS

C eed nd — m

z 0 0 2 8$ 000072 0 00 0 I 0 ! & , 2
£L 0 1|) Z 068 0186€ zz € 0 ¢ OF | 4 WH v X
0 0 0 0 000000 0 00 0 0 0 0 0 ^
0 ao € 000072 0 0 0 0 02 GE tb 68 60l A
8 Ii Il vb b | € € Æg 0 Z €8 OP #9 SII E83 H
6b 00 6 88 S 8 £e g 69 æ I 1 4 Obl Zbl 88 bbl 49b s
88 9] cl Æ I | gt 4 v9 2 Et 10 NI S8 Il el Sb | 1
0 +r 0 Z? 00 | 020 0 00 6 S 0 S J£ b
€ iW 6 S 0 0 0 0$ eo 0 2 4 Lf 09 Z6 Zl d o
y 0) S B Z b cc || 68 G9 C 0 € % 001 £S% 991 zoe u ja
6 9 Ə € 00 001] ıı b 0 I| OL SS IZ @ 612 u P
86 B € v 1 8 8 II Gt 5 O ll 02 Zel ch €2I gIE fakes
0 2 0 | 0 60 0 O10 0 00 2 ¢ 0 8 œ ft |)
$ v 0 8 00000 g 0 00 € 8 c G 9 u
0 @ 0 06 000010 0 00 ol U Z3 Ze Gg 8
0 tm o H 00 0 Ot 0 O 00 € AZ wl WE ££ J
S £8 | +t 1 60 | 1 zz O I E 4S & 61 I8 £8 P
€ SG | sl 0 90 02 9 0 | v SG Sh tL £9 s9 5
0 8 * 1! 000000 0 00 6 & Se 2 OF q
m cct SEE HERE EN
00 0 0 0 00001 ! 0 00 0 * 0 I| 0 1
L € H H 00 001 * 0 00 0 I € Z9 A
4 $66 Ol OL L € OI € Zo G 6 00 p 89 OS! 06 99I n =
4| OOI SS Ze! S 8 ZI zt eb Gb £l b 1 89 > BH 91 vil log
09 £Z IZ & Of 2 Z vl GI t4 GE 0 € os! 8110 0 9 o i-
221 991 cL £I 8 S6 Le t€ I8 €9 ZZ | & 06 A 0 Of 29 e
¿ZI Z0£ 6i% BIE O€ Ob Cb £E £83 89] OF 0 9 9E HI 9 29 96 ə

d u u | f 4 B | p 23 q 3X A n |») 0 ? a

sjuruosuog o
eS
youalg *(71002) g*p 318 VL

158

MESSAGE DECODING

IV.

000 0 00 000 0 0 00060 0 0060 0 0 10. 0 I I 0 I z
000 0 0 0 06 0 0 0 000 000 10 0000 1 0 4 x
0 09 0 WF 0l ? oz 8 st + 099 6 2 at t 61 0 8% tS o 16 9 ^
000 0 4 @ 0&8! ! € 0000 01! 0 0 £ 0 1 & og VL Ii A
0 0 Ib ZL W $ |» £c bh L| Ot Gt 0 2 GG 06 cc ti | S 3g GEL IZI OL] O8 OZI s
0 00! Z Sb w | O L| Tr Z4 L 0 €& I SI vb vl Bt | ce TE 4 GIT col Gel ale 1
oor 0 I |! 000 00 000 F 0] 00 0160 0 8 1-0 I| b
0 S0 ef & (0C o 9:0 9 æ 108 1 * S I ! Sg hH 6i W9 bE Ot 19 d o
0 0% | b L| O 0 (OC S 9 9 o zc WS Scl 9] II | E £9 GIT €0c piZ SZI vgl u 3
008 ft Z4 Ww 09 9 & 8 60 08 P 6&0 1 Y gt bI 9 8» 28 OL 601 w o
0 ost € oO 2 0 ÆN 8 Æ flor Z 4669 8 £L Zv 86 tb 88 POI 6lI Gel pe
001] 0 S8 2 0O IL 99 0 & 00€ I $0 UT 20 2 €% Of BÆ SE y 5
000 0 0 0 000 0 0000 0908 0 I 0€ rt 0 0 0 I f
0.099 0 V €& 0 8 4 Sr € 0A Sb Gl 6b Ob 9 06 6! GE OFT 26 16 0gb u
006 0 S E | I! 91 6 Z? los 9 & BT € s € 0; 6€ v9 Ib 96 ig
0021. 0 0 8 0 P S Z 4 6 Oc * 9 H it ! 8 L SE ze 2 16 2G 365
0 07 1 6 P | S SIO £9 0 cz ob GI Zzgl SI AL tl 09 £L 68 IL 0B P
0 LI 0 € Wl 0 | 9 ! 8 TT 0% | LS 0 0 ¢ 6 tl 9% Ob 66 c9 27
0 01 0 €& 8 0 ! UH 9 4 !? 19 € ISO 8l æ 8 S PG 65 OL q.:
1 06 £ Z t€ | S tl OF cb G 00g 9 8 1c B |6 O E8 WV» LI GG A
000 0 @ IE zl + £9 H gZ O eZ vE L tl 6 GE | 0 0 4 6 r MB n
0 0 8& | cel ZL O 6 GIT OT €h Z4 1 we OG Gc 09 El 8 ZE 1g Ol ZOI gel ZSI 801 1
E 0 €$ € I&I 6I 8 #9 800 Gb 88 £Z 0 Ol GE 4€ EL 9 S 8 6 2010 3 0% £9 EG
I 0 06 OE Of SOI | FE FIZ 78 FOI 90 O 26 t9 22 68 OF W th b El Sb 9 GI ZN els
0 0 16 PI 08 Gel O Of Scl OZ 6l 8€ O I6 Ih 26 14 GE 65 Li SIT ZS] OG SI FIT 8b oz
I 4 901 Ut OLI ZIE I 19 PST GOI sel Ge | Otb 9c ZS HOT S9 OL S8 BOL C9 BIT Sb 9lI a
27 X M A s 1 b d u w 1 4 f— y 8 y p 3 q A nm |) !» e o ð

STOMOA

S1u?euosuo:)

usr2u3 :(1002) e'p 3TEV.IL

159

EXTRATERRESTRIAL CIVILIZATIONS

Since lost their vocality in the living language. The French letter k occurs
very seldom, mainly in abbreviations (e.g., in initials of non-French names).
The error in the English-language table is associated with the use of the

letter "t" in combinations which represent a distinct phonetic sound. An
algorithm correcting errors of this kind and leading to successful results is
described in /12/.

One of the "mathematically correct" algorithms minimizing K, is given on
p.187. A related algorithm which converts the so-called "syllabic writing"
into normal letter writing is given on p. 188, and an algorithm identifying
classes of words with a common meaning is described on p. 192.

$6. MATCHING ALGORITHMS (PART I)

Algorithms identifying code sequences

Algorithms intended for the detection of larger textual units, when the
smaller elements are known, evidently constitute one of the most important
classes of recognition algorithms.

We start our description of these algorithms with one of the simplest: an
algorithm identifying letter codes by uniform-length sequences of symbols.

The importance of this problem for the case of extraterrestrial communi-
cation is obvious. The "elementary signal" of a message transmitted by an
extraterrestrial civilization may have a simple form, in particular repre-
sentable as one of the two binary symbols, 0 and l. To transmit a longer
alphabet, coding will have to be used, representing letters by sequences of
the elementary signals. These signals quite likely may be of uniform length
for all the letters of the alphabet.

The set of permissible solutions (the set of interpretations) in this case is
found without difficulty. Let m be the length of the code groups, and N the
length of the text expressed in elementary symbols. The number of permis-
sible solutions in this case is m: it is determined by the number of shifts of
the text through i digits (i= 0, 1, ..., m—1). Cyclic arrangement is
assumed, whereby the last letter of the text is followed by the first letter.

If N and m are relatively prime numbers, the residue obtained in the
division of N by m will be omitted. Therefore, the total number of permis-

Nm
Sible solutions is X [=] , where [=] is the whole part of the corresponding
m=!

quotient. This number is not greater than

Let us now proceed with a discussion of the quality function. Consider a
text of length N encoded by groups of numbers of length m. What distin-
guishes this text from a random number sequence partitioned into blocks of
the same length m? It is obvious that the frequencies of the m-letter groups
in the second case should be much more uniform than in the first case. After

160

IV. MESSAGE DECODING

all the second number sequence has been picked up "at random," and none
of the numbers has any preference over other numbers.

On the other hand, the selection of letters in an ordinary message is far
from random. There are sounds and sequences of sounds which are relative-
ly easy to pronounce; if the message alphabet is the set of words in the
message, different words occur with different frequencies, because of
considerations of common usage" and depending on the meaning of the
message,

If the encoded text contains groups of length m, and we attempt to interpret
them as containing groups of length p(p#m) or at least groups of length m but
displaced through i positions (where ij and m are relatively prime), the
message becomes similar to the sequence of symbols obtained by "repeating
selection." This means that the elements of the code group corresponding to
a single letter are more intimately related than the elements which belong to
different code groups.

This sounds reasonable because incorrect "partitioning into groups" is
devoid of those "preference criteria" which restricted the letter combinations.

It is therefore natural to use a quality function which reaches an extremum
for a uniform distribution of the code element frequencies and also for a
certain "highly" nonuniform frequency distribution.

Unfortunately, intuitive reasoning is not enough for an a priori choice of
a quality function assessing diversity.

A whole range of traditional evaluation techniques are known. These
include, for instance, the calculation of the root mean square deviation, the
modulus variance, the entropy.

Our calculations based on a limited text pointed in favor of the function

y -X6)-90.

Here cj is a certain group of a given length, (c) is the mean absolute fre-
quency of a group of this length, equal to APO where m is the chain

length, N is the length of the text in unit symbols, |A] is the number of letters
in the alphabet of unit symbols, |A|» is the number of letters in the alphabet

of groups of length m, x is the number of groups of length m in the given

text (rounded off); thus V is the sum of the squares of the deviations of the
actual frequencies from the mean absolute frequencies of the groups.

For a given group length, the function V is minimum when all the group
frequencies are equal (then V = 0) and maximum when one symbol recurs
through the entire text. ]

Topermit comparison of the results for various group lengths, the expression

X(9()- pe?

is multiplied by a normalizing factor v. This factor can be calculated if we
proceed from the assumption that the best (maximum) value of V should be
independent of m. Since the best (from the point of view of the particular

161

EXTRATERRESTRIAL CIVILIZATIONS

function) solution involves a single element only, its frequency is m and
the frequencies of the other elements are zero. Then V is equal to
N N N 2 m4) —
(siam) siam) UAP- D=
QNT (1 oe il ie ar |e
E il yam) — gam? mam? O?
N?

and since usually |A|" is large, we may take |Al"-1«1A/l", so that V ma ~ P
Let the group length in some other solution be /; the maximum value of

2
V is then approximately equal to NN.
The normalizing factor is introduced so that the best values are equal:

N? N?
MY
Hence
12
ven

In the example that follows, a partition into groups of length m — 3 is used
as the "basis for comparison"; for 3-digit groups, the coefficient v is equal

to 1.
A short English text* has been encoded by a sequence of three-digit

numbers using the following table:

a = 000 j=100 s = 200
b —001 k - 101 t=201
c = 002 (= 102 u = 202
d=010 m= 110 02210
e=011 n-—lll w=211
f=012 o=112 x= 212
g=020 p= 120 y = 220
h=021 q= 121 z-291
i= 022 r= 122

The last three-digit group is not used. The encoded text will look as
follows:

201 021 O11 201 211 022 020 021 102 022 020 021

201 022 200 200 000 010 000 111 010 002 102

112 202 010 220 201 021 011 211 022 111 010

001 102 112 211 200 211 022 102 010 000 11! 010

012 122 011 011 000 111 010 000 200 201 021 O11

211 022 111 020 200 112 012 200 O11 000 001 022

122 010 200 012 102 000 200 021 201 021 011 211

021 022 201 O11 002 000 120 200 112 012 201 021 O11 200 O11 000

* The first stanza of R. L. Stevenson's poem "Twilight":
The twilight is sad and cloudy,
The wind blows wild and free,
And as the wings of sea birds
Flash the white caps of the sea.

162

IV. MESSAGE DECODING

Table 4.4 lists the absolute frequencies of the three-digit groups for
partitions arte with the first, second, and third letter of the text,
respectively (R3, R3, and R3).

TABLE 4.4

R? 200 011 000 021 010 022 a 211 102 [11 013 uz 020 001
10 9 9 8 8 6 5 5 3 2

RÈ 110 010 000 001 002 100 112 210 020 122 220 221 021 202
2115 8 7 6 6 6 6 6 5 4 4 4 3 3

R? 102 101 020 000 201 001 011 120 229 100 121 211 002 202
3|14 011 9 7 6 85 5 5 5 4 4 4 3 3
R? 002 122 120 202 220 109 101 110 121 210 212 221 222

[ 2 2 1 1 10000000 0

R3 102 120 121 200 212 222 011 012 211 022 101 11! 201
2 2 2 2 2 2 | I 1 000 0

210 110 200 212 012 021 022 221 O10 111 112 122 222
3 2 2 2 1 |! 1 10000 Q0

The factor vy is taken equal to unity. We find
V (Ri) 2340.30; V(R))-277.34, V (R3) 321.37.

The absolute frequencies for two-digit groups and partitions beginning
with the first (Ri) and the second (R3) letter of the text are given in Table 4.5.

TABLE 4.5
| 10 | 00 | 20 [or | n [oo | n | 22 | 12
| 28 | 24] 19 | 17] is faa | io | n | 9
o2 | o0 | or | 11 [ao | 20 [m | 12 | 22
Re 2 |: [m | iz [is ie [io |n] e 6
92
Taking for the normalizing factor v = $; = 0.44, we find V(R?) = 268.00-

a
-0.44— 117.92, V(R3)= 295.44 0.44 = 129.99. Both figures are markedly
less than v (RÀ), The absolute frequencies for m = 1 are

9(0)-124; p(l)=94; o(2)—

The normalizing factor is equal to n we thus have V (RI) - 1176- $- 130.67.

This is again less than V (RÌ).
Consider the absolute frequencies for four-digit groups. In this case, the
partitions may start with the first, second, third, and fourth letter of the

text (RÍ, Ri, Ri, RÍ, respectively).

163

EXTRATERRESTRIAL CIVILIZATIONS

The frequencies of the symbols are listed in Table 4.6.

TABLE 4.6
A 2 P A 2
29|.£& eo}. & obja f vba l
Sslos 11538 ys (i| SE os 4/38 oz

A" |S$ER ES "7 $2 E&|^3 $8 ES| sissies
Oo} su. A| su AGI Sw agian
<a rao <i 14 0 <t1Z 0 <E IZ 36
s | 1 5 | 2 4 | l 4| 2
a] 2 a| 5 3 | 9 3a |7
3 | 4 2 | 12 2| 9 2 | 48
2 | 12 1 | 24 NE" 1 | 2
NE o | 38 o | se o | 35
AER | |

2
1
The normalizing factor is E =~ 1.67, and V(Ri)= 165.77, V(R)=

= 169.02, V (R$) = 141.38, V(Ri)= 125.88.

To establish that V is indeed maximum for the correct partition, we
Should calculate the values of this function for all m (m —1;2;,5.5 i).

This is, however, not absolutely essential: clearly, the squares of the
differences ( (ci) —9(c))? markedly decrease as m increases, whereas v
increases only moderately.

Another thought is that groups longer than twenty elements need not be
considered altogether; after all, even assuming a binary set of elementary
symbols, the power of the alphabet of groups of this length is 279 i.e., more
than enough to represent the most complex alphabets (including the Chinese).

Let us compare the entropies calculated for some of these partitions.

The entropy H = — Ñ, p; log p; for a uniform distribution is maximum; it is
i

zero if the probability of one element is 1 and of all the others 0. We

replaced the probability p with the modified quantity p=260 , where z

is the length of text in terms of m -digit groups.
We have H(R)) = 1.165, H(R2)= 1.248, H(R3)= 1.226, which is again quite

satisfactory. For all other lengths of code groups, the entropy should be
normalized by dividing by log m. It is remarkable that the V corresponding
to almost all incorrect partitions have close values: this again proves the

adequacy of normalization.

An example illustrating the application
of the concept of meaningfulness

The above algorithm can be applied to demonstrate the concept of mean-
ingfulness, presented in $2. Let us identify the set of interpretations of the

164

IV. MESSAGE DECODING

text with the set of permissible solutions of the given logarithm, defining
meaningfulness as the difference of the functions V or H for the worst and
the best partition.

The results which can be obtained following this approach are best
illustrated by an example. Consider a binary alphabet of elementary signals,
with two elements 0 and 1. The code group is 2 elements long. In this case,
the set of interpretations contains only two partitions, Ri with the groups
starting with the odd elements, and R$ with the groups starting with the even
elements.

How do these interpretations look in the best case, when A(H)= H (Ri) -

— H(R$))is maximum? Let us characterize the partition by the probability
distributions of the code groups:

Partition
| RÌ OR

Distributions

These probabilities are not independent. They can be expressed in terms
of the probabilities of 4-digit groups beginning, say, with elements whose
running number in the text is a multiple of 4:

p’ (00) = p (0000) + p (0001) + p (0010) + p (0011)
p” (01) = p (0100) + p (0101) + p (0110) + p (0111)
p' (10) =p (1000) + p (1001) + p (1010) + p (1011)
p’ (1) =p (1100) +p (1101) + p (1110) +p (1111)
p" (00) = p (0000) + p (0100) + p (1000) + p (1100)
p” (01) = p (0001) + p (0101) + p (1001) + p (1101)
p” (10) = p (0010) + p (0110) + p (1010) + p (1110)
p" (11) =p (0011) +p (0111) +p (1011) -- p( 1111)

Inserting in the expression for A(H) the values of p'and p” expressed in
terms of p(aaaa), we obtain a function of 16 variables, whose maximum
will enable us to compute the two distributions. The distributions [p;) and
{p/) characterize a text of maximum meaningfulness in a certain sense!
Pit br

Taking averages of the form p?= , we obtain the probability distribu-

tion of the pairs of elementary symbols of the text, and taking the sums
p(01)+ p(00) and p(01)+ p(11) we obtain the probability distribution of the
one-digit symbols (pf?). These distributions permit computation of the second -
order entropy of the text using the equality H,— H (pf?) — H (P), and hence the
approximate redundancy (or, more precisely, the lower limit of redundancy)
as 1— Ne replacing Hæ with H, and Hy with log 2.

Highly interesting results could be obtained if the "intelligibility" were
evaluated in terms of entropy of infinite order. Calculations carried out
under this assumption would help to estimate the level of entropy and
redundancy characteristic of meaningful texts.

165

EXTRATERRESTRIAL CIVILIZATIONS

$7. PATTERN DECODING ALGORITHMS

The language of images. Connectedness and detailedness

Our definition of intelligibility is neither widespread nor usual. Usually,
a text is considered intelligible if it produces mental association with some
real situations or images.

This approach naturally does not answer the question why certain situa-
tions from reality are unintelligible. Nevertheless, the usual interpretation
of intelligibility is largely valid. After all, there is a fundamental correla-
tion between the "predictive systems" of the human language and the human
reality, or, to use a different turn of phrase, words combine roughly in the
same manner as the real phenomena that they represent.

Thus if man speaks, moves, and interacts with the surrounding objects,
the word "man" will naturally also combine with words designating speech,
motion, action. Another remarkable correspondence is observed between
sentences, which are generally made up of words designating objects (nouns),
actions (verbs), and properties (adjectives), and typical real situations
which are made up of objects, their interactions, and properties.

This correspondence is far from trivial, and yet it is not too complicated,
so that no special translation rules had to be devised in any of the languages
for one particular situation.

The translation from the language of reality to human language and back
is naturally a very complicated undertaking; there is, however, one peculiar
human language for which this translation is done without much difficulty.

We mean here the language of images.

This language clearly suffers from considerable shortcomings. It is
highly uneconomic: e.g., compare the sentence "a man walks" with a picture
announcing the same fact. A correct image must contain a great wealth of
detail, which is often immaterial for the case being considered. Moreover,
Some messages do not lend themselves to translation into the language of
images without sacrificing the simplicity of the mapping which relates the
image to the real situation (e.g., such sentences as "perseverance wins" or
"1963 was a droughty year").

There is therefore no reason to suggest that the language of images would
be the only means of interstellar communication.

However, its great advantage is its intelligibility. It is not only that the
image language can be readily translated into the usual language of reality:
there is a very strong predictive relationship between the adjacent elements
of an image.

The graphic form of the decoding problems associated with image analysis
is another highly favorable feature, enabling us to consider these problems
as models for the more difficult task of decoding of the ordinary language.

One of the typical decoding problems in the analysis of image languages is
the following: consider a sequence of signals; it is required to convert it into
a two-dimensional picture so that an intelligible image message is obtained.
A typical feature of this problem is that it gives rise to serious doubts
concerning the usefulness of a formal definition of "meaningfulness." After
all the human mind will immediately distinguish between a meaningful and
a meaningless picture.

166

IV. MESSAGE DECODING

Let the sequence of signals comprise the lines of a rectangular scan of
an image consisting of black and white dots (represented by the digits 1 and
0, respectively) arranged in succession. 'This sequence can be decoded in
the following way: by changing some parameter d from 1 to N (N is the
Sequence length), we partition the sequence into lines of length d and arrange
these lines one under the other.

A man examining the picture formed in the process will instantly
recognize the best of the various images.

The reader may wish to experiment on his own with the following
cosmogram:

10011111001001111111000000001
00

11111110000011100000001111100000
1
10010111110100101111101001011111
0
11111010010111110100101111101001
0
10000110110000001101100000011]011
0
11011000000110110000001101100000
0
0010011011001

The decoding of this message is

10011111001

00111111100
00010101000
00011111000
00001110000
01111111110
01011111010
01011111010
01011111010
01011111010
01011111010
01011111010
00011011000
00011011000
00011011000
00011011000
00011011000
00011011000
10011011901

and the ones form the picture of a man in a hat.

This method of image decoding on a rectangular screen will possibly be
the most effective if a computer is entrusted with the task of deriving the set
of all possible solutions (i.e., the partition of the text into segments of
length / and their arrangement one under the other). The situation radically

lo7

EXTRATERRESTRIAL CIVILIZATIONS

changes, however, if we consider messages obtained by scanning a screen
of an arbitrary shape. In this case, the number of permissible solutions
increases prohibitively.

Even a computer will not be able to examine and assess all the permis-
sible solutions in this case. However, if the computer has been programmed
with a formal criterion of meaningfulness, it may apply a shorter procedure
proceeding, say, from a somewhat less meaningful image or part of an
image to a more meaningful one. The difference in meaningfulness between
the two successive images may be so slight as to be actually imperceptible
to the human eye.

Consider two image scanning techniques: the image is covered by two
systems of x and y coordinate lines, and the lines of each system are
spaced a certain distance Ap. The part of the image between two adjacent
lines.of the system x will be called a line, and the part of the image between
two successive lines of the system y will be called a column. The element
of the text located between adjacent lines of the systems x and y will be
called a dot. If a certain classification of the dots is given, each dot may
be replaced by the corresponding classification digit; the lines of digits are
then numbered and arranged in sequence one after the other.

A preliminary hypothesis for the construction of an image quality criterion
presupposes that the system of coordinate lines always can be defined in an
optimum manner for the particular image.

BECCIITEETIEETTETTETTITTITETETT TEETTCETTETEE TELTETTCETETTTTTTEIETTET TETTE ETT [ETT ITTTHES

FIGURE 54. In this picture, rich in vertical details.
the “horizontal lines” are highly similar.

In this, we have to lean on a certain property of texts, which is apparently
fairly general, Meaningful messages probably fall into component parts, not
unlike sentences, which in their turn are composed of smaller elements
(analogous to words). The "quasiwords" in the "quasisentences' should be
different in a certain sense; the adjacent "quasisentences,' on the other hand,
should be close to each other in a sense.

Thus, Russian-language sentences are made up of words expressing a
variety of concepts: nouns signifying objects, verbs signifying action,
processes, or states, adjectives qualifying properties. The real images or

168

IV. MESSAGE DECODING

Situations corresponding to the different words in a sentence are also highly
different. At the same time, nearby sentences have similar structure and
often close meaning, which imparts the sense of "connectedness" to the text.
For instance, the sentences "A vessel emerged from beyond the horizon.
This was a boat with a wide white stack" deal with a common subject, ex-
pressed by the words "vessel" and "boat," and they thus appear "connected."

These properties are even more prominent in a picture: if Figure 54 is cut
into horizontallines, the black and the white dots will frequently alternate;
adjacent lines will moreover be very similar to each other, whereas in
columns the black and white dots alternate infrequently.

cu "Py j A
Baa IP) | (as) (CO

FIGURE 55. Pictures rich in horizontal, radial, and concenrric lines.

This is clearly true only if the partition into lines and columns is done
according to a certain pattern: thus, for the picture of a crocodile the lines
should be vertical and the columns horizontal; for the picture of a flower,
the lines are concentric circles, and for the picture of an apple they are
radial lines (Figure 55).

A similar property is characteristic of messages composed in formal
languages and in LINCOS-type languages: adjacent sentences in these
languages are "logically sequential" and they are generally similar to one
another when presented in graphical form.

For fairly complex images, the choice of correct coordinate lines is
apparently not so significant, because for any direction of the lines, two
adjacent lines will have a similar appearance and will contain a frequent
alternation of black and white dots. We will say that similarity of adjacent
Scanning lines ensures connectedness of images, similarity of more distant
lines ensures smoothness, and variety within the lines ensures detailedness.

Examples of quality functions. Some procedures

A simple quality function can be proposed evaluating images in terms of
connectedness and detailedness. Detailedness is assessed as the number of
transitions from a black dot to a white dot within a single line, and connec-
tedness as the number of black— white transitions occurring in corresponding
positions in two adjacent lines. Let 1 stand for a black dot and 0 for a white
dot. The function u; i}, assessing the quality of adjacent lines can be written
in the form

Uis = 9(101) + o (110),

where g(|01) is the number of transitions from a white to a black dot in the
(i+ 1)-th line occurring below identical transitions in the i-th line, q(|10) is

169

EXTRATERRESTRIAL CIVILIZATIONS

the number of such binary transitions from a black to a white dot. The
line-wise image quality is expressed in the form

Uiine= x Ut tet

Since we do not know in advance if the "quasisentences" are lines or
columns, column-wise quality function should also be evaluated, identifying
closeness of adjacent columns according to the formula

u; 477 (01) + e(T0),

where 9 (01) is the number of white-to-black transitions along the vertical,
situated next to the corresponding transitions in the column immediately to
the left. The column-wise image quality is then expressed by the equality

Usum 2d up pi’
and the overall image quality U is given by

U-Zu, ict ur, uu (4.4)

The summation can be carried out along the lines only, since the sum

2 ui j, 18 equal to the sum J Wii Where ti pa — 9 (01) + p (T0) in adjacent lines.

It is not entirely clear how to treat the first and the last symbol in
adjacent lines. In our conception, the partition into lines is equivalent to
introduction of special "boundary symbols." The transition to a boundary
symbol naturally carries certain information and should affect the image
quality if it is transmitted together with the image. However, since by
assumption the image being decoded does not contain special boundary
symbols, this information is "fictitious" and should be minimized.

FIGURE 56. Thepictureof an inclined solid line obtained
with a square grid.

For simplicity, we will assume that the image is unbounded in both the
vertical and the horizontal direction, i.e., it is drawn on a torus (a steering-

170

IV. MESSAGE DECODING

wheel) the last line is followed by the first line, and the right- most column
is followed by the left- most column. If the last line is partially filled, it is
completed with a more frequent element, e.g., with zeros.

Consider some of the first decodings of the text

00000010100101001 110000100001000000

with the respective quality functions (for the other decodings, only the quality
function is given):

Length of line 1 2 3 4 5 6 7
0 00 000 0000 00000 000000 0000001
0 00 000 0010 01010 101001 0100101
0 00 101 1001 OIOIO 010011 0011100
0 10 001 0100 01110 100001 0010000
0 10 010 1110 00010 0000010 1000000
0 Ot Oli 0001 00010 000000
1 01 100 0000 00000 U=2 Us6
0 00 (001 1000 7-10
I 11 000 0000
0 l0 100U—4
0 00 000
1 01 000
0 00Uz4
1 00
0 10
0 00

Interpretation 1 00
1 00
1 U=6
0
0
0
0
1
0
0
0
0
I
0
0
0
0
0
0

U=0

Length 789 10 11 [2 13 14 15 16 17 18 19 20 21 22 23 24 25
of line
64260644 84182484844 4

Length of linc 26 27 28 29 30 31 32 33 34 35
U 4444444440

A correct interpretation is the decoding with a five-element line (the
picture of the numeral 4). This line length also corresponds to the maximum
value of the quality function U — 10.

The above quality function is suitable for images rich in thin and solid
vertical or horizontallines. However, it will give erroneous results for
discontinuous images, and for images with prevalent diagonal lines. Both
cases are interrelated; indeed, a square grid cannot form a continuous
image of a diagonal line (while preserving the line width). Let us examine
Figure 56.

171

EXTRATERRESTRIAL CIVILIZATIONS

In this figure, allthe centers of the black squares lie along the straight
line y — ue xt i It is readily seen that not a single additional square can
e

be hatched without breaking this condition, The quality function avoiding
this difficulty makes use of what is known as image smoothness, We define
a special operation, called "linear forecasting," whichassignsa third line A,
to any pair of lines A; and À;.

This operation is carried out as follows:

The elements of the line 4; are joined with the elements of the line 2; by
Straight segments observing the following three conditions:

1. Every element of the line 4, is joined at least with one element of the
line Àj.

2, Every element of the line A; is joined at least with one element of the
line Ai.

3. When conditions 1 and 2 are observed, the sum of the segment lengths
is minimum.

The segments are then continued to an arbitrary distance. If the segments
are continued to row j+k=/, we say that the maximum forecast depth is k.
A forecast of depth & is implemented as follows: the squares with the
Segments passing through their centers are identified as black squares (ones)
and all the other squares remain white (zeros). For a rectangular screen,
the length of the line qp is the same as the length of the lines q, and q; (in
our example, & —1) (Figure 57).

In other than rectangular screens, the position of the boundary points is
first determined (Figure 58).

g 91001101000 ROTEL d
EON ` 10+

g O0L0LL0T%00 bt al fae
Nord gS 3 i

Gf oanr11 0100 FR I

FIGURE 57. Forecast of FIGURE 58. Forecast of

depth 1 on a rectangular depth 1 on a screen of

screen. arbitrary shape.

The resulting line is then compared with a real line 4; occupying the same
position, using the function u, ,.

In practice, it is probably always sufficient to compare two adjacent lines
and to make a forecast of depth 1, i.e., to forecast the adjacent line.

If the number u,: obtained from a forecast using the lines A; and Ais: is
designated uw, (Ay i); the image quality may be estimated with the function

Use ^E A u, t (L, tu).

Example. For the pattern
0000000
0100010
0010100
0001000
0010100
0100010
0000000

172

IV. MESSAGE DECODING

Uine =10, which is much better than Uis = 0.
Similarly to linear forecasting, we could define nonlinear forecasting,
which uses three lines to reconstruct a fourth. For example,

0100001000
0100001000
0010000100
0000100001

Here the difference between two adjacent horizontal shifts is preserved.

'These techniques of bypassing the difficulties associated with discontinuity
are logically irreproachable and do not look excessively arbitrary. A sim-
pler method is described in the following.

The presence of diagonal lines with squares touching at the corners may
be allowed for if the lines in (5.4) are replaced with bottom-to-top diagonals
and the columns with top-to-bottom diagonals. Designating a pair of transi-
tions along adjacent diagonal "lines" as /01 or /10, and a pair of analogous
transitions in the diagonal "columns" as N01 or X10, we may define the

similarity of diagonal lines and columns as u%8 — 9 (/10)-- 9 (/10) and wj°7% =

= 9 (101) + @(\10) and the "diagonal quality function" as U% = D up% + D uti.

The total image quality is expressed as the sum of the two "quality"
functions:

ue = U+ Use,

For fairly complex images, however, the function wis quite sufficient.

a) ii iii di dH 11_ 4417 "ue per “ive
1 7 7 8
111 1 1 4 2 2 3
1 1 1 1 4 4 g 4
1 11.1 1 4 1 4 o 4
111 1 1 2 2 2
2 2 2
1 1 1 ao g g
11 1 Oo 7 7
11 11 11 11 11 1 Ü & 5
11 111,1 11 1 1 a 5 5
11 1 1 11 11 11 11 11 Ü 5 5
a? md 35 2 ; 2
gg
111 1 1 11111 1 i P Aia 2 2 2
g
11 14.1 1 2 7 3
111111 1 1 11 1 1 Z7 2 3
1 1 1 1 1 d 2 2
11111 11111 1 2 4 8
1111 111 111 1 7 7 2
1 11 11 4 o 2 2
1 1 1 1 1 1114 111 6 2 à
1111 1 1 1111 1 1 5 3 9
1 1 1 1 1 1 1 1 2 Z 13
111 11111 111 1 e 6 a
1111 1 111 11 1 1 1 7 5 8
1 1 111 1 1 1 3 3 8
1 1 1 1 1 1 1 1 2 2 4
11 1 1 1 1 1 1 1 13 7 74
1 11 11 11 11 11 11 111 68 é 8

V tine column. “V2
-56 45

FIGURE 59. Drake's cosmogram. The correct interpretation (line length

41). The first line is preceded by the last line and the last column is

followed by the first column.

173

EXTRATERRESTRIAL CIVILIZATIONS

Figure 59 and Table 4.7 show Drake's cosmogram. The values of
4, qa; M and the corresponding sums are also given. Note the consider-
able difference in the values of u for the correct and incorrect interpretation.
Both pictures are assumed to be drawn on a torus.

TABLE 4.7. Drake's cosmogram. Incorrect interpretation (line length 64). The quality (142) is much lower
than the quality of the correct interpretation (172).

Up je Up jua Hp qe Up py]
0100000 1000001 1000000001 1000001 101 10001101 10000011001 1 1000000000 0 ll ll

1000000000000000000000000000000000000000 100001110000000000000100 1 9 2 2
00000000 1000 1000000001000 100000000000000000000000000000000000010 0 0 0 0
0010000100000010000000000001000000000001000100000100010010010000 0 0 2 2
010001000100000001!100000000000000100000000000010000000000000000 0 0 2 2
0000000000000000000000000000000000000000000000000001000000000010 O 0 2 2
001000001 1000 10000000000000000000000000000000000000000000001 1000 0 3 1 4
01100001 100001 100001 10000100000000010010010010010010010010010010 0 4 5 9
0101010010010000 1 100001100001 100001 100001 1000000000100000000000! 0 3 6 9
1111010000000000000000000001000000000001000001000000000001011011 1 l 7 8
100 100000000000001 1111010000000000000000000000000000100000000000 1 0 7 7
0000001000100111000000000000101000000000000000101001000011001010 0 2 5 7
1110010100000000000000010100100001000000000010010000000000000000 1 | 5 6
9100100000 1000000000001 11110000000000000111110000001110101000000 0 4 8 12
1010100000000000 101010000000 100000000000100010100000000010100010 1 8 4 12
0000000000000000010001001000100010011011001110110110100000100010 0 4 0 4
0010101010001000100000000000000000010001000100100100010001000000 0 4 li 15
10000000000001 1 1000001111100000111000000011111010000010101000001 1 4 12 16
100000 10001000000 10000000000 100000100001 11000010000010000011000 0 5 1 6
000000 1000001000100010001000001000001000011000010000010001000100 0 4 4 8
01000001000001 1000000001 1000001101100011011000001100111000000000 0 - - -
U 47 U. 95

column

For fairly large images, it is difficult to try all the possible alternative
linelengths, especially in manual work. We will therefore propose a less
reliable, and yet much faster method. The same method has been applied
for image decoding on a screen of an arbitrary shape, when it is in principle
impossible to examine all the alternatives.

If a boundary element is interposed between two elements, its position
from some initial point u can be identified as i. We will say that the point
n; has a U-neighborhood if on both sides of the boundary symbol in position i
from the origin there are segments d such that when the right-hand segment
is placed under the left-hand segment, we obtain two lines M, À for which
U atu, >U,

Let the set of points with a U-neighborhood be {p;(U)}. The simplest
procedure using the properties of J-neighborhoods is based on the assump-
tion that there exists Uma: such that (u;(Umax)) * 1 (the power of the set of
points with a U-neighborhood is 1). This is interpreted as follows: the
image contains a pair of best lines, which are very close to each other and
pass through the part of the image rich in details. Thus, (nu (Umax)) contains
a single point H e» The length of the U -neighborhood for Hpe is the length
of the line; Hype itself is the reference point.

The following procedure therefore can be applied: setting U= 1, 2, 3, ...
we establish whether or not a particular value of U is attained for more than

174

IV. MESSAGE DECODING

a single point. If this is so, U is increased by unity, and the search is
repeated. Otherwise, we have located a single point with a U-neighborhood.
Given this point (u,,,), we use the length of its neighborhood to determine
the length of the line; the position of the point identifies the beginning (the
end) of the lines. If there is no such point, the line length is identified with
the length of the U-neighborhood of one of the points with a (U-1)-neighbor-
hood. In this case, the solution is not single-valued.

Example. Decode the pattern

010100101001110000100001.

1) Is there more than one point wi for which U 2.1? The answer is yes,
e.g., u2, Hu.

2) Is there more than one point for which U2 2? Again yes, e.g., the
Same points u2, n7.

3) Is there more than one point for which U2 3? No, there is one point
us for which Umax= 4; the length of the line is 5.

The answer to the decoding problem is the outline of the numeral 4 on a
rectangular screen,

When decoding patterns on a screen of arbitrary shape, u,,, is obtained
according to the same rules; one line is then added from above and one from
below to the selected pair of lines, whose length and position are chosen to
ensure a maximum increment in U. Changing the beginning and the end of
the adjacent lines, without altering the value of U, we ensure maximum
smoothness of the boundaries, using one of the proposed functions.

Example. Consider the message

0000001 1111000000000001000 10000000001 1 1 1 100000001010100000
10101000111110000000000.

All the points u,,,, lie densely between position 53 and position 58 (this is
a property of messages on an arbitrary screen). The pair of lines corre-
sponding to one of the points is given below:

0000010101
0000010101

For these lines U;,.= 6.
The best position of the next line is the following:

0000010101
0000010101
00011111

Here Us 3= 2.
Construction of the next lines does not alter the value of U. The lines
stacked on top are

00000011111
0000000000010001
00000000011 1110

00000010101

175

EXTRA TERRESTRIAL CIVILIZA TIONS

and on the whole

00000011111
0000000000010001
000000000111110
0000010101
0000010101
00011111

which gives the correct answer: the pattern of a "window." The original
pattern is

00000011111000000
000001000100000
0000111110000
00010101000
001010100
0111110
00000
000
0

i.e., a "window" on a triangular screen. Thus, although knowledge of the
Screen geometry is essential, it does not alter the pattern itself.

$8. ALGORITHMS ANALOGOUS TO ALGORITHMS
WHICH CONSTRUCT BILINGUAL DICTIONARIES

Letter-comparison algorithms using the properties
of close neighborhoods

In the previous sections we described examples of various algorithms
analyzing texts written in an unknown language.

We have mentioned before that the aim of this analysis is to construct
the best interpretation containing information needed for the most effective
forecasting of the inaccessible part of the text for any arbitrary partition of
the text into accessible and inaccessible pai'ts.

Suppose that such an interpretation has been found. Examining the accessible
part of the text, we willbe in a position to predict what comes next. We will
possibly learn to construct "correct sentences" or even "correct texts" in
the new language.

The next question, however, is concerned with a more fundamental aspect:
is this really what we sought to achieve when we started the decoding? The
answer is an emphatic no. After all, we still do not know how to translate
the message into a known language and into the "language of reality." In
other words, to give a crude example, we still cannot build the machine that
the message describes.

To effectively translate an unknown text, we should establish a corre-
Spondence between the elements of our language and some elements of the

176

IV, MESSAGE DECODING

code message. The corresponding elements in either language may be
selected by a variety of techniques; for example, we can compile a list of
sentences in the unknown language and their translations into our language;
or we may assemble lists of words with the appropriate word translations.
The best and the most natural approach is probably to compare certain
linguistic phenomena on which the predicate system of the two languages is
based. We have mentioned previously that the basis for the analysis of
textual meaning is provided by the "semantic classes of words." A bilingual
dictionary with ordinary words replaced by names of semantic classes would
be shorter and better than a conventional bilingual dictionary; a dictionary of
sentences, on the other hand, is impracticable and cannot be drawn up even
for a pair of known languages.

Translation from one language into another thus requires bilingual
dictionaries of certain elementary phenomena which make the text. This is
a necessary condition, but obviously insufficient. We should, moreover, be
able to compare the rules according to which the elements of the two lan-
guages combine betweenthemselves. Afterall, the same words canbeusedto
give sentences with entirely different meanings.

In other words, we should be able to define the "closeness relation' in
the two languages and, when preparing the translation, we have to ensure
that the words of the translation are represented by the same closeness
relations as the words of the source text.

The decoding algorithm, however, is never expected to produce a polished
and styled translation. It is quite enough if the algorithm provides sufficient
information for a human operator to prepare the finished translation. The
development of a "dictionary" and "comparative grammar” is therefore one
of the last aims of decoding algorithms.

We will describe an algorithm which compiles a dictionary of sorts, but
the component elements of this dictionary are letters, rather than words or
semantic units. The starting assumption is that a certain "anthropomorphic"
(i.e., vocal) language is to be decoded and translated into another known
human language. We know how the letters of the known language are pro-
nounced, but the pronunciation of the letters in the other language is unknown.
Our aim is to describe the pronunciation of the letters of the unknown language
using the letters of the known language. In the simplest case, this can be
achieved by establishing a "correct" one-to-one correspondence between the
letters of the unknown language and those of the known language.

We will describe a simole algorithm which establishes this correspondence
or substitution. We will also consider certain means for finding more
general solutions.

The set of permissibie solutions in this particular case is the set of all
one-to-one mappings of the letters of the unknown language into the letters
of the known language. If the two alphabets are of unequal length, the
smaller of the two should be supplemented with an appropriate number of
letters which are interpreted as "accidentally missing" from the text.

The basic hypothesis used in the construction of the quality criterion for
the letter substitutions is that letters conveying similar sounds should have
similar "combination properties." A particular form of the algorithm
depends on the precise interpretation of the concept of "combination
properties."

177

EXTRATERRESTRIAL CIVILIZATIONS

A similar assumption is naturally introduced when compiling dictionaries
of "semantic multipliers"* and words. It is assumed that elements of
similar significance or meaning follow the same combination pattern.
Although in certain cases this assumption is not absolutely obvious, it
nevertheless provides the only conceivable basis for decoding.

In this algorithm, the combination properties of the letters will be
represented by a table T of the frequency of occurrence of the various
letter pairs. The number P(a, aj) at the intersection of row i and column j
is expressed by the equality P(a,, CEA ere where a, and a; are
the letters of the given alphabet, ọ(a; a;) is the number of times the pair of
letters a; and a; occurs in the text, N is the length of the text.

The combination properties of a letter a; are defined by the row of
numbers of the form P(a;, ax) in the table T.

The dissimilarity of two letters a; and a; in the same language can be
measured using one of the equations expressing distance between the points
of an n-diraensional space.

We define the following measure of dissimilarity of the letters a; and aj:

n
o (a,, a)-2 | P (ai, a4) — P (aj, az) I,

where a, a, a4 are letters of the given alphabet.
Now consider a certain substitution x: A— B of the unknown alphabet A into

the known alphabet B. Let aj be the image of the letter a, under the substitu-
tion xn. Then the function

G(a,, aj, n= > |P (ap a,)— P (a; a,)|

characterizes the dissimilarity (with regard to combination properties) of the
letter a; and its image a; under the substitution a.

The quality of the substitution x can be found as the sum of dissimilarities
of the letter pairs entering this substitution, provided their dissimilarity is
evaluated for the same substitution. For the quality p of the substitution we
thus have

p= X 9 (a,, a;, n). (Z)

in]
We should thus look for the best substitution 1,,,, such that
p(m,,)- min.

In principle, x,,, could be found by examining all the possible substitutions
and estimating the quality p of each using equation (Z). However, the number
of possible substitutions is very large (n!), and we will therefore propose an
abbreviated procedure which, we hope, is equivalent to the complete proce-
dure for all practical purposes.

We shall say that x; is a transposition of s;, or n; (m;) if to some pairs
a, —a;, a,— a; of the substitution n; correspond the pairs a,-»a;, a,— a; of
the substitution nj, whereas the other pairs of the two substitutions coincide.

* i.e., names of classes of semantically close words.

178

IV. MESSAGE DECODING

As is known, for any two substitutions n, and z,, we can always construct
a sequence of substitutions zs ..., 44 ..., mı, where m, is a transposition of
Tw; nxis a transposition of nu, and for any pair m; mili <x; i51), misa
transposition of m;.,. The sought substitution is therefore always attainable
by advancing from any given substitution through a number of intermediate
substitutions, as indicated.

The transposition x; of some substitution nz; may be a best substitution
only if p(x;) €&p(zx)). If for some s; and any x; (xj) we have p(xj) > plm), we
say that the function p has a local minimum at the point m;. Clearly, if p
has an absolute minimum at the point 3,,,, it also has a local minimum at
the same point Apes-

Therefore we can find the absolute minimum z,,, by examining all the
possible local minima.

We cannot propose a suitable method at this stage, but we will describe
a procedure which locates a sufficiently deep local minimum.

For some substitution n;, we define the set of transpositions (zx; (m)}.
For each transposition x; (z;) we calculate the increment Ap, equal to
p(x;) —p(x;). Then we choose s, such that Ap(n,) is maximum and positive
in the set of the increments (Ap(z;(1)))). A similar procedure is repeated
for {n:(mg)}. The routine ends when we have found a substitution for which
no transpositions with positive increments exist.

To calculate the increment Ap, we have to determine p(n,) and p(z;). If
7", differs from s, in that n; contains the pairs a,—»2,0,—*a; and a, the pairs
a, =a, a, >al, the increment p(n,) — p(n;) can be calculated using the following
Somewhat cumbersome formula:

2 2| P(a,, a )— P (a; aj)|-2 | P (as, a.)— P (ai, aj)|-
-|P (2. &.) P (ar. a7)| - | (as a.) - P (0; a)]—
= 22|P(a. a,) - P (a 22)|-2 2| P (a, a,)—P (a, a7)| -
*- | P (a, a,) — P (a; a?)| — | P (a; aj — P (a; af)|.

As the initial substitution, we can choose the one that is obtained when the
two alphabets are arranged one next to the other in the order of decreasing
numbers,

ZP (a, a,) and XP (aj, a‘),

respectively.

The initial substitution will include pairs of letters a, — a, which have the
same number i in the corresponding sequences,

We will now summarize the algorithm in the form of a system of general-
ized instructions.

1. For the text to be translated, draw up a table T, of numbers

9 (az aj) + 9 (aj ai)

P (a,, aj) = 2N ,

where f(a,a;) is the number of times the pair a,a, occurs in the text, ~N is the
length of the text.

179

EXTRATERRESTRIAL CIVILIZATIONS

2. Draw up a similar table T; for the text in the known language.
3. Is the number of rows in 7, and 7; the same? If not, complete the
smaller table with dummy rows consisting of zeros until the two tables are

of the same size.
4. Arrange the rows and the columns of the two tables in the order of

decreasing numbers 2 P (a, a) and DP (a, a’).
5. Construct all the Aina) possible pairs of the form a; a; and for each

pair calculate the increment Ap(a;, 2j) —Ap.

6. Are there any positive Ap(aj, aj) ? If none, proceed to instruction 7;
otherwise, go on to instruction 9.

7. Print out the answer: the set of pairs a,» a;.

8. End.

9. Interchange row and column aj with row and column a, in the second
table for which Ap (aj, aj) is maximum; return to 5.

In the case of combination properties represented by an asymmetric

Har aj) , we should proceed along

matrix with entries of the form P (a; aj) —

the same lines. The only difference in this case is that the increments A
have to be calculated using a different formula.

Let us now briefly describe an algorithm using a more general trans-
formation.

We are looking for a transformation A which meets the following require-
ments.

1. For any letter a, (a;&4A), there is a pair aa; ell.

2. For any letter a; (a; €B), there is a pair a, a; ell.

3. To each of the alphabets, at least one "dummy" letter at is added,
such that p(aj, a,) = 0 for any ax.

4. When conditions 1—3 are observed, the sum of the numbers P (a, aj)
is minimum.

Thus, a mapping of the form

a a, 2, 4, as
g & E Q
is permissible, while the mapping
Kd G 45 4 a
G BG G da a

is unacceptable: here the pair aa, can be omitted without breaking conditions
1 through 3.

The distinctive feature of an algorithm using this mapping is that the
length of each row in tables 7, and Tis doubled, and this in addition to the
elementary manipulations which interchange the rows of the second table.

The elementary manipulations can be divided into induced and free.
Indeed, the result of some manipulation may be an unacceptable mapping.

To ensure an acceptable mapping, some additional elementary manipulations
Should be carried out.

180

IV. MESSAGE DECODING

Free elementary manipulations are those which are not intended to
restore the mapping to an acceptable form. The true gain of an elementary
manipulation will be defined as the gain of the elementary manipulation
proper plus the sum of the gains of the best elementary manipulations
induced by the particular elementary manipulation and restoring the mapping
to an acceptable form.

At every step of the routine, we have to calculate the true gain of all the
free elementary manipulations. "Then if the gain is positive, we have to
carry out the entire sequence of elementary manipulations corresponding to
the truly best free elementary manipulation. The routine is terminated
when the true gains of all the free elementary manipulations become negative.

Computer experiments were carried out using this algorithm. English
and French texts of 5000 letters each were selected for this purpose.

The correspondence is shown in Figure 60.

etoransadriwertmugbpuvuytrsyg
Peed ddd dd deeded ddd ddd
esaniturtbtod@eopmuvgsgthjrzywe

FIGURE 60. A substitution obtained in a computer experiment analyzing
the correspondence of Fnglish and French letters,

Tables 4.8 and 4.9 contain the relative frequencies of two letter sequences
in English and French. For convenience, all the numbers have been multi-
plied by a factor of 10,000. The order of letters in the corresponding
Sequences was ignored.

The rows (and columns) with numbers of the same order of magnitude
should be close to one another, since the tables have been processed by the
algorithm. We must confess that the similarity between the corresponding
rows is not very striking.

The results of the comparison are evidently quite unsatisfactory. In any
case, they are no better than what could be obtained by a simple frequency
analysis of letters.

Let us try to establish the reasons for the poor correspondence.

In our opinion, the main reason is to be sought in distortions of ortho-
graphic origin. Thus, the two-letter combination "th" conveys a single
sound in English; the letter "y" in English is sometimes pronounced as a
vowel and sometimes as a consonant. It is surprising, however, that the
same effect has not distorted the vowel and consonant identification procedure.

The inadequate results of the computer experiment focus our attention on
the tremendous difficulties in the direct comparison of the linguistic elements
of different languages.

We will now show that a preliminary analysis markedly improves the
quality of the comparison procedure. Let us first divide the letters of both
languages into vowels and consonants (this algorithm, as we know, yields
virtually error-free results), and then compare tables corresponding to the
two classes V (vowels) and C (consonants).

181

EXTRATERRESTRIAL CIVILIZATIONS

———M——HÁÁÀÁHÉÁÉÉÓÉMÁÓ ai aiaa
————————

oel

086
OFI
OPI
Ort
Stl
061
607
022
006
00S
osz
668
S8e
ose
FOL

oes

6901 6S0l P62

0000 0 0 0 0 0 0 G 0 0 S S S 0 0 0
0000 S 0 0 0 S 00:0 0 0 0 0.0 ¢ 0
0000 0 0 0 0&0 0 0 0 0 0 S 0 E 0 0
00086 0 0 0 60 0 0 6 0 S 0 0 0 S1 0 0
0S8 00 0 OL OF op 08 Oc St OS GH w sI oS OF 06 OF
000000 0.0 0 S 0 0 S or o 0 StI SI 02
0000 010 0 s OQS O80 Qt 0 Qo VHH 9 g
00 00 010 S O10 S Of 0. 09 O S S S&I Ost OE
0900 00 0 o OF SI W St GI GG; g 0 OF SIT SI
0000 0S O c SI Ob olt OG SOI Oc sc 01 00€ OL GE
$6, 000 S$ S10 S Q S8 OILO S O8! OS OS S OF o6z OF
0000 090 0 og St 06 W OF Oc Oc s oz 09% c8 OZ
000S S6lS OL og SOL SOI 08I OG O6h OPI 98 of S&P SZ SIZ
80000 OS 01 0 or Se O06 OS OG Obl OE] OC Or SSI ost S9
0038350 St 0 09 ¢ S S6 08 S SB O8 oF 0 Sig n6 Ge
S 000 00 9 g¢ 0 S g 06 Ot OF O 09 Slá O9 06
0 0 c£ SI Or Sel G6 sop OF ODE OF O42 SEP GG] GIG Glo 0 see SSE
0 ¢$ 00 06 SI SI Ost SII OL 06% S8 Gz OS! 06 09 S6t ozi S
0 0 0 0 OF 02 06 O GZ SE Oh OL SIZ S9 ce 06 SSt cpi 06
S 010 O0 (06 S6 GL OFF 09 OS 08% Ov! Obi Ob] svi O61 6027 022 002
0 0 s Ol OF OF OHN SI SE Ote SZI SI SE Se SEI Olr GOL! SZI 618
0 010 Ol SEI OEI SS 06% O11 OGI S6 STE Sor SOI OPE 686 PZZ FEO Sos
000 0 Sb S O SG OI 09 09 SG OF SOI GIG Sez OES Gk] SIZ
09 0 00 OOI SZ OEI OPI SZI OI] Olp SIE SOh 6E9 SOZ See SG] G9 023
0 S 02% 081 0 SE SG OP O8 GIE S6 YZI Ste 08 GLI 664 Sze 966
S SI Sz OSZ OFZ SPE OL! 04% OLE Occ O8 SIS S09 S956 Oh OtG Oto t8ÞI
bf x 2 Á ^ JX) d q 38 n u | J 2 “oi 1 p

+08
FLS

0 0

0 S

0 0

0 or

Sv ool
S SZ
0€ oer
S6 — OVI
ot SZI
095 OIT
09 olk
S ste
OP — Sor
gor Oto
SIZ 606
S&B oge
oes 66I
Sbi ¥s9
SIZ 0/6
SBE OSE
OCI +39
6S9 68

0, Sot
Soe OFF
8106 GFZ
€4l6 O9T

0 S b
E ST r
0c 92 x
S Go z
Oc; O£6 LS
0 She A
S£ OZI J
ss 046 d
Or — Ott q
08 oez 8
sie Sz n
$6 SIS w
zı #09 1
4 J
08 0cb 9
S41 «ORS ^
6hL OEZ !
Set FBP

+
e
o
bed
D
us
ow Oc ww fF w^»'S.

usrigug "8'p 318V.L

182

MESSAGE DECODING

IV.

coocoo0o0o0000000000€€oo0o0o0oooocococooocoooococ

"oco 00000505»0»00005

Oocw»coooo0nocoooococoooooooocoo
eg

S
€

0 0
0 0
0 0
0 0
0 0
0 S
S 0
0 0
0 si
0 s
0] 60
s Ol
o 0l
S sI
Sı 0€
oE o
s Ol
03 S6
0 OI
S 08
S $6
Sg! S8
0 $
o 0€
S OI
99] $8

0.0000 0
0.0000 9
0.0000 0
0s 00 0 0
¢ 0 0 81 S S
0.0000 0
0 0 0 0 0 0
o 0 06 0 0 0
0. 0 0 S 0 S
0.00000
0.006 0 0
0.0000 0
0.00 O 9 S
0 00 0 0 0
0 660 OO SI
o 0898 0 S
09 OL gg SS OG sS
€ S 09 0] S8 99
st 0 ge 06 Ol OS
sg € o¢ OF Of SIS
s; 0 02 S Æ 09

S $9 c8 S6 00I 06
0g Sc og S»! Sl OF
Sc Sel 08I SSI ShI GG
Sb ST 04, SE 09 08
GL} 06] SZI SIC O91 SII

e wu
— N

cOw»ocococococoocsoc

S86

uou
a=

S01

Obi

egg

vco

0 0
0 0
S s
oe s
o ol
0 09
€ o
0| cg
S gs
0 o
S g
0 cg
Sí «0e
SI cp
0 osr
OS og
08 o
OF oog
00% szt
00t 666
08! OEE
S0£ GO
SIE 688
ooe SI
OFZ Ske
6ZI1 SZ

0 0

0 0
se OF
06 0
& 0l
S €i
S1 0
09 SE
Or 06
s8 on
S9 0€
S! S6
Si s
S41 S06
os S06
Or 006
008 Scb
OLb 013
oiz OPI
08% 6/9
00€ 008
e9 628
OI? S8
GF] $89
04£ SIE
8441 869]

HS
08€
619
v001

081 Soe
oee SOP
00E t9
ozs 6/6
S8t 0S6
046 +901
¥90! 06
6r9 SZS
6/8 +36
V989 6v8

Sii

SER

£6

bE $29 89/1 082

0 0 x
0 or ^
06 o Á
S S9I z
0t $8 x
St gw il
sl 060 |y
$9 s |J
S cuc 3
09) 09 q
08 ci b
SL 69 ^
c6 veg w
SM Gg d
ssl 666 12
OZ Gef s P
sze o8 o
oe gi ' 1
sig 969] 1
619 $00] n
S99 wel a
68 veo o
cee 89] u
v6 085 è r
6911 £900 $
£905 Iz >?

youaly 67b 38V.

183

EXTRATERRESTRIAL CIVILIZATIONS

V C

Since the numerical entries in the off-diagonal blocks VC and CV are
substantially greater than those in the diagonal blocks VV and CC, it suffices
to calculate the sums in VV and CC. We have*

VV cc
French text 1972 4302
English text 1330 5261

We conclude that the first class of French letters (vowels) corresponds
to the first class of English letters. We further divide the vowels of each
language into subclasses:

French text:
e,0,0 uy

English text:
e O a,i, u, Y

a, uy

A comparison of these tables establishes a correspondence between the
French e, a, o and the English e, o, the French i, u, y and the English a, i, u, y.
Corresponding subdivision of the consonants gives for the French text

Srn, dl, m, h, x, z (class 1)
b, d, g, p, UG E c q, fu, w jo (class 2)

and for the English text

S, r,n,l, m,h. x, q, j (class)
b, d, g, pt, k, c, f, w, v — (class 2)

The diagonal squares of the corresponding tables are the following:

cl.1 cl.1 cl.2 cl.2
French 1174 416
English 936 372

We have again obtained a correct correspondence. Further subdivision
into smaller categories in principle could provide detailed information on the
correspondence of the individualletters of thetwo languages, but unfortunately,

^ The calculations were carried out using Table 4.3.

184

IV. MESSAGE DECODING

for fairly small texts, the information concerning combination and frequency
properties is too scanty.

The next algorithm possibly can be applied using the preliminary division
into classes and restricting the comparison to letters which belong to the
corresponding classes only.

The above calculations show that in order to establish a valid corre-
spondence, the "predictive systems" of the text should first be analyzed. In
particular, it is clear beyond all doubt that straightforward comparison of
letters in the two languages is doomed to failure. It is the "semantic
classes' obtained by a separate treatment that should be compared.

An algorithm using distant neighborhoods

The previous algorithms represent the so-called "statistical" approach
of the military deciphering techniques. There are algorithms, however,
which utilize the alternative conception, namely the method of characteristic
words. Let us consider the set of permissible solutions and the quality
function of such an algorithm.

This algorithm also establishes a correspondence expressed as a certain
substitution. Let us first estimate the quality of the pair a, a, entering the
substitution 3. We will designate the letter a, which occurs in position p in
the unknown text by the symbol a.(p), and the letter a which occurs in
position ¢ in the unknown text by the symbol a; (t). A pair of letter sequences
will be called a permissible u, v -neighborhood of the pair a,(p) a(t) if for
every k(k< uj and m(m >v) the pair a, (p +k), a, (t+ k) and the pair a,(p — m).
a,(t—m) are elements of x. The difference u — v will be called the length of
the permissible u, v-neighborhood. The maximum length among all the
permissible u, v -neighborhoods of the pair a, a; will be used as the quality
of the pair a, a;, or q(a, ai). The quality Q of the substitution x can be
defined as the sum of the qualities of the constituent pairs:

Q(x) - 2o (a,, a2).

The higher Q(x) the better is the substitution. For the best substitution
Keer» Q (ana) is minimum.

While the previous algorithms of this section used the properties of
nearest neighborhoods, the present algorithm is based on the similarity of
long letter sequences. Both principles naturally can be combined into a
Single procedure.

$9. CLASSIFICATION ALGORITHMS (END)

"Mathematically" correct algorithm for vowel-and-

consonant identification

We will describe one of the "mathematically" correct algorithms
minimizing the function K;.

185

EXTRATERRESTRIAL CIVILIZA TIONS

We will say that K, has a local minimum for a partition R=), £y of the
alphabet A into two classes, if a transfer of any letter from the class k, into
the other class kand vice versa does not generate a partition kj, k, for
which K, is smaller than for R=k;, ko.

If there is no partition such that K; is smaller than for R, we say that the
function K, has an absolute minimum for the partition R.

Clearly, if the function has an absolute minimum for the partition R, it
also has a local minimum for that partition. Therefore, the absolute mini-
mum can be found by examining all the local minima (i.e., all the partitions
for which K, has a local minimum).

A subset & (f c A) is said to be permissible if for every a, (a, & 8) we have

> (a1 ae) — P 9 (a, aj) 7 0.

Here m—|$8], a, ER; a,e$8; a, ANE.

The following two theorems hold true:

Theorem 1. If both classes of partition &,, k, are permissible, the
function K, has a local minimum for &,, ko.

Suppose that the proposition of the theorem is not true, i.e., there exists
a letter a, which reduces the value of K, when transferred from k, into kz.
The classification obtained when a, is transferred from &, into k is desig-
nated kj, kj, so that ki, k;— 5, Na, EU a,. The value of K, for kj, k, is

n n

> Š plan aj) + X A P (as, a) —

ol {mt kom+l lem

^|
~

—2 2 ola a)+2 D Plax a).
i=l komt!
The value of K, for &, & is

> 2i q (a, apt > > , 906 a).

i=l j =m+l lemt+

By assumption, we have

à p> g(a: aj)-- res x 9 (a4, a)— » Sea, aj—

= È X , P(e a)+2 i 9 (a, a;) io,

komt!) l=m+
n

=2 x 902 a4) 2 0

k=m+

m n
p 9 (az, a) — = 9 (ax, a5) >0,
i-l kam+!l

which is impossible, since a, belongs to a permissible class and thus
satisfies the inequality

n m
È (a, a) - D glar a) 7 0.
kamtl i=l

IV. MESSAGE DECODING

Theorem 2, Ifthe subset &, is impermissible, any subset &, contain-
ing &, is also impermissible.
Since &, is impermissible, there exists a letter a,, such that

>} play, ar) xa 2 o (a, ai) < 0,

k=m+l

where a; a, are elements of &,, and a, is not an element of this subset.
Consider the following expression:

t
X i 9 (as, a,) — 2 9 (ax, as),

ral+

where |8,|—/; a, is naturally an element of R, a,c&8,a,€& AN S,. Itis
readily seen that

n t
> e (as, a,)= TOS (ay, Q4) — 2. plar, ay),

r=f+l
where a,G&,\8,, and hence

>» glay, a)< X 9 (ax, ar)

r=l+l m+t

at the same time

H m 1
Dolan a)= Boar, at D olay, aj).
s-1 tat p=m+l
t m
so that 2 @(a,, a) is greater than or equal to Bolan aj). Al the more so,
$^ -

n Li
a 9 (ay, a,) eS P 9 (a,, as) <

ae, n m
< X glas a,) D glas aj),
k-2mt*l isl

and it is thus negative.

The two theorems lead to a construction of an algorithm which examines
all the permissible classes. Theorem 2 enables us to shorten the examina-
tion procedure, We then seek all the possible nonintersecting pairs of
permissible classes which cover the entire alphabet. Every one of these
pairs is a local minimum in virtue of Theorem 1. The absolute minimum is
picked out from among the local minima by direct examination of the alter-
natives. Let us summarize this algorithm in instruction form:

1. Construct all the possible subsets of the alphabet, containing ¢ letters
each.

2. Omit those subsets which are not permissible.

3. Draw up a list of all subsets containing £ + 1 letters each and not
containing any t-letter impermissible subsets.

187

EXTRATERRESTRIAL CIVILIZA TIONS

4. Is this list empty? Yes: proceed to instruction 5. No: substitute
t+ 1 for ¢ and return to instruction 2.

5. Find all pairs of nonintersecting permissible subsets covering the
entire alphabet, compute the value of K, for each pair, choose the pair for
which this value is minimum; the particular pair of classes provides the
solution.

6. End.

The initial value of the parameter fis 2. Theorem 2 helps us with
instruction 3 of the routine: in virtue of the theorem, we do not have to
construct all the possible subsets containing t+ 1 letters each, but only
those which are made up of permissible /-letter subsets. The permissibility
check is based on the inequality

È elas, a) - D lay, a) >0,
k=om+) i=]

which should hold true for every letter of a permissible subset.

An algorithm translating syllabic writing
into alphabet writing

The vowel-and-consonant identification algorithm assigns the value of one
binary feature to each textual element. The algorithm discussed in this
subsection, on the other hand, assigns the values of two features to each
element, and these are moreover multidigit, and not binary, features.

This algorithm is evidently far from being able to determine on its own
the number of distinctive features and their possible values. It nevertheless
has a two-fold practical importance. First, it may help to establish the
pronunciation of letters in the so-called syllabic writing, and second, an
analogous algorithm will be useful for morpheme identification.

There are many examples of writing in which a single element corresponds
to a sequence of sounds, rather than to a single sound. When a syllabic text
of this kind is to be decoded, it should first be transcribed into normal alpha-
bet writing. Then the pronunciation and the grammar are easier to determine.

The sequence of sounds corresponding to a syllabic element is often not a
true syllable. It commonly has the following standard structure: the first
sound is consonantal and the second vocalic.

This, in particular, is the case in Creto-Mycenaean writing. For example

b-ty, T-£5 Fx M -pu, 5-70
p-ti, E-za, M-va, l-ve , etc.

In decoding such syllabic writing, the algorithm should translate each
syllabic symbol into a pair of symbols, whereby the second symbol is
common for all syllabics with the same main vocalic sound, and the first
symbol is common for all the syllabics with the same consonantal sound.

The set of these symbols constitutes an "ordinary" alphabet replacing
the syllabary.

The morphological interpretation of the algorithm will be discussed
later on.

188

IV. MESSAGE DECODING

The algorithm is built as follows. Two classifications k, and & of the
syllabary S are built. According to the first classification, syllabics with
a common vocalic part, e.g., ta, pa, ka, etc. are grouped in one class;
according to the second classification syllabics with a common consonantal

part (fa, tu, to, ..., etc.) are in one class,
The classes from k, are identified by the symbols a, %2, ..., and the
classes from k, by the symbols Bi, Bo, ... . If k, is a vocalic classification,

each syllabic can be assigned a sequence of the form fi, «,. An ordinary
alphabet A corresponding to the syllabary comprises the symbols a, 02, ...,
Gm, Bis -> Bae The alphabet A may also contain symbols for "null" vowels
and consonants. A ""nullvowel," i.e., a symbol denoting the absence of

a vowel, is required to transcribe a syllabic which corresponds to an
unpaired constant, and a''nullconsonant' is required for the transcription
of vowel syllabics.

It turns out that &, and & cannot be defined arbitrarily. The sought pair
of classifications should have a certain restrictive property. If a given
pair has this property, we will refer to it as a permissible pair.

Having defined the set of permissible pairs, we examine it for the best
pair, i.e., the one extremizing some quality function.

Let us describe a permissible pair of classifications. It is readily seen
that if one classification contains m classes, each class of the other classi-
fication will contain at most m elements. Suppose that k, comprises two
classes, the class of syllabics corresponding to the vocalic sound a and the
class of syllabics corresponding to the vocalic sound i. Consider the class
of syllabics in &,which contain the consonant p, say. It will naturally
contain the two symbols pa and pi, or only one of these syllabics (since some
elements of the syllabary S may be missing from the text). No third element
in addition to pa and pi may enter this class, since there is no third vowel
to combine with the characteristic consonant of the class.

In some syllabic languages, different syllabics may convey the same
sound sequence, but then our algorithm is inapplicable.

It is moreover assumed that pairs of classifications for which fewer
syllabics are missing from the text are relatively more likely to be valid.

Let us consider the quality criterion to be applied to a permissible pair.
Consider a table 7, =|{f(s;s;)l| where the entry in row i and column j is the
number of ordered groups of the form sis, occurring in the text. The row i
of the table therefore contains numbers which indicate which elements of S
follow some given s; and with what frequency. Let s; be decoded as Bx, a.
It would naturally seem that the appearance of the row i depends to a greater
extent on a, than on f, and therefore rows corresponding to combinations
ending with o, should be close to one another. Conversely, close columns
are those which correspond to groups beginning with the same letter.

Closeness of rows and columns can be estimated in terms of some
distance between the points of an n-dimensional space (where n= ISI ) We
will use the simple relations

n
9, (s;, s)7 D1 olsise) — e(,s)l for rows, and

189

EXTRATERRESTRIAL CIVILIZATIONS

9; (Si, s) = 211 (ses) — e (s5))I for columns.

Here o; and o indicate dissimilarity (distance), and s, runs through the
entire syllabary.

We can now construct two tables T, and T; presenting the row and the
column distance of the table T,.

Let the adjacent rows in T; and T; correspond to the same class, so that
the table can be divided into strips, each containing rows of one class only.
The tables will take the form shown in Figure 61.

The hatched squares in Figure 61 contain entries which indicate the
distance of elements of a certain class from other elements of the same
class. In virtue of our basic assumption, these distances
Should be small, and the sum of the numbers in all these
Squares will also be relatively small if the classification
is likely.

We designate the sum of entries in the quasidiagonal
squares of the table 7) by X,, and the corresponding sum
for the table T, by X,!. To construct a quality function, we
use the obvious line of reasoning, which maintains that in
a good pair of classifications, even the worst classification
is sufficiently good. Therefore, we may take

FIGURE 61. Partition
of a table into strips
corresponding to vo- Kalki, ka) = max (2,, X).
calic classes.
Here K, is the quality of a permissible pair. Unfortu-

nately, the meaning of K, is clear only when the number
of elements in each class of one of the classifications is precisely equal to
the number of classes of the other classification. In our more general case,
we either have to use a more complicated function or to alter the definition
of a permissible pair. We adopted the second course.

The final definition of a permissible pair is therefore the following:

1. If one of the classifications contains m classes, the number of elements
in any of the classes of the other classification is at most m.

2. The sum of the distances from an element s; to other elements of the
same class is less than the sum of the distances from s; to elements of any
other class.

3. With conditions 1 and 2 satisfied, no other pair of classifications ki
and k, exists which satisfy conditions 1 and 2 and such that

[ei] 9 1&il and [e| 14.

The simplest search algorithm for a permissible pair of classifications
calls for a detailed examination of all the possible alternatives, rejecting
the impermissible ones and picking out among the remainder the one with
minimum K,. This algorithm, however, is quite impracticable.

We will therefore propose a shorter routine for deriving the best solution
which, we hope, is equivalent to the complete examination procedure for all
practical purposes.

We construct two sequences of classifications k, ki Rig, ... and ka, Roo,

Table T; corresponds to one of the sequences, and T; to the other.

190

IV. MESSAGE DECODING

Using one of the tables, we proceed to construct progressively finer
classifications, whereas the other table is used to construct progressively
coarser classifications. The sums Zi, decrease, while the sums Y;
increase.

The value of K, is first determined by the sum F, and then by Y,, as
Shown in Figure 62.

1
2

FIGURE 62. The functions X, 9 and K4.

The thick line in the figure is the plot of K,. We see from the figure that
the minimum of the function occurs at the point x, i.e., the sequence of the
classification pairs can be terminated when we reach a pair of classifications
built on the basis of rules 1 through 4 in which Èa; is greater than X.

We will not go into a more detailed description of the computation proce-
dure. It suffices to note that it largely draws upon the algorithm described
on p. 155.

Our algorithm may be given an interesting twist if it is applied to words,
rather than syllabics. Each word may be regarded as an element of some
"grammatic' classification and an element of some "lexic" classification.
For example, the Russian word "3owoit" [home, in the sense of "going home']
is classified in one grammatic class withthe words "ctosom'!, "020i, "peiuenmiewtt,
[declensions of the words "table," "water," "solution"] in that they are all
instrumental case singular. On the other hand, this word is part of another
class containing the words 103r" '10x2a', '1032x', etc., all based on the same
root "10oĪm" [house or home]. The relationship to one of the grammatic
classes in Russian is generally identified by the suffix. This is not always
so, however, and the nominative case has no identifying suffixes. In this
case we are dealing with "null suffixes."

The entire situation is completely analogous to the various assumptions
regarding the structure of syllabics. Similar considerations can therefore
apply to the construction of a quality function.

The table 7, can be replaced with a table of conditional probabilities that
if a word A, appears in a simple sentence, thenthe word 4, will also appear
in the same sentence.

"Lexically" similar words shouid "control" other words according to the
same pattern. "Grammatically" similar words, on the other hand, are
conversely "controlled" according to the same pattern: e.g., the accusative
case is conditioned by the presence of the so-called transitive verbs in the
sentence. This is entirely analogous to the starting assumptions used in the
construction of the quality function of the algorithm.

This method of analysis is fairly interesting in that it detects "null"
morphemes. At a later stage, we will consider a morpheme-identifying
algorithm which is unable to identify words. This algorithm, however, will
not identify the "null" morphemes, either.

191

EXTRATERRESTRIAL CIVILIZATIONS

The algorithms probably should be applied in the following sequence:
first, we detect the "non-null" morphemes, then words as certain combina-
tions of these "non-null" morphemes, and the words are then again parti-
tioned into morphemes by the above algorithm.

An algorithm for "semantic" classification of words

Classification algorithms are particularly significant for decoding the
meaning of a text.

We have mentioned before that external dissimilarity in the appearance
of words does not always indicate that the words are significantly different
in their meaning. The verbs "to have" and "to possess" are quite dissimilar
morphologically, and yet their meaning is evidently very close. It is
generally difficult to predict which of the two words, "car" or "automobile,"
will be used in a sentence, but they are evidently synonymous.

A classification grouping words of similar meaning in an unknown text
will greatly enhance the intelligibility of the text. This operation is there-
fore a necessary step in any decoding routine. As we shall see, a ''seman-
tic" classification also provides a quantitative estimate of the "semantic"
closeness of words.

We will describe a method proposed by Yu. A. Shreider for developing a
system of classifications. In this method, binary and ternary classifications
are assumed, i.e., classifications partitioning the entire set of words into
two or three classes. For a binary classification, one ofthe classes contains
all the words with a certain common semantic feature, and the other class
comprises all the remaining words. In aternary classification, there is
another class of words with an opposite property. Thus, one class ofa
binary classification may contain words associated with the concept of
"space," whereas the other class will contain all other words, unrelated
to this concept. The concept of "animation" may constitute the basis of a
ternary classification: animate, inanimate, and words to which this
feature is inapplicable (e. g., the word "show '').

A partial list of useful distinctive features (classifications) is given
below: 1) intelligence, 2) elementarity (the property of being unique,

3) action, 4) animation, 5) positiveness (good — bad), 6) greatness (large —
small) 7) space, 8) time, 9) order, 10) the property of being a boundary,
11) perception, 12) change, 13) the property of being a part of a whole.

The presence of a certain feature in a given word will be identified
by the appropriate number, and the presence of the opposite property will
be identified by the number with a bar on top. Some examples of semantic
codes of various words are given below:

algorithm 3, 9, 12.
computation proces: 3, 9, 12.
instant 6, 8. 13.

fool 1, 2, 4, 5.

rest 3, 5, 8, 13.

French plural article 2.

192

IV. MESSAGE DECODING

The "semantic codes" can be replaced with vectors, thirteen-
dimensional in this case, with the numerals 1, —1, or 0 in the -th
position, accordingas the particular word has the property in question,
the opposite property, and no such property:

fool 1,1,0,1, —1,0,0,0,0,0,0,0,0.

For this thirteen-item semantic classification, closeness is estimated
using the equality

ex (a, 5) = D Llen ny),

where e; is the i-th coordinate of the word a, qj is the i-th coordinate
of the word 5,

0, if e;—m. ]
1, if g-2zlmw--l i

mS , 4.5

Hen n)7]15 if 0 n=l, | A
| or e 7 £l, m=0, |

Here the distance is chosen so that words with opposite properties are
closer to each other than words characterized by presence and absence
of a property (e.g., the words "giant" and "dwarf' are closer than the
words "giant" and "philosopher"). The distance px (a, b) is independent of
the text: it is determined by the choice of the semantic categories.

A certain textual distance is defined in the following. Using this
distance, we can establish a list of categories (the set of classifications).

Let the distance L(a, b) between the words a and bin a sentence be
defined as the number of words from a to b, inclusive. (We know that
visual distance does not always correlate with distance in meaning. The
distance is therefore measured using the so-called graph of the sentence,
and not the straightforward text.) Thus L(a,a)= 0, and for adjacent words
L(a,b)= 1. The distance is defined as the mean value of L(a, b) for all
sentences with a and b occurring simultaneously. If the words aand b
do not occur in a single sentence in a text T, we take pL(a, b) — oo.

The function p}(a, b) characterizes the distance between "combining"
words, whose semantic relation is such that they appear jointly in one
sentence. For "mutually exclusive' words, such as "house" and "hut,"
the function p!(a, b) takes on very large values, despite the obvious
closeness in meaning of the two words. Mathematically, this shortcoming
of the function p!(a, b) is manifested in the fact that it does not have one of
the basic properties of a metric distance: the "triangle inequality"

(p(a, 6) - p(b, c) Z p(a, c)) is not satisfied for this function. We therefore
have to define the distance pr (a, b).
Let

eta, 5) = min [(p} (a, c) +p} (e, 5).

p$ la, b) = min [o$ (e, c) + e" (e, 5)]. (4.6)

EXTRATERRESTRIAL CIVILIZATIONS
The textual distance pr(a, b) is defined as

0; (a, b) - lim pr (a, b).

This quantity satisfies all the properties of a normal distance.

Given a certain set of classifications, we can apply pr(a, b) to examine
a certain text and hence to improve the original set of categories. This
is based on the assumption that the distances pr(a, b) and px(a, b) are
consistent. For example, we may assume that a small pz (a, b) leads to a
small px(a, 6), and vice versa.

We will now describe the application of the algorithm. The semantic
vectors are defined for a list of words picked out from a given text
(Shreider suggests assigning semantic codes to predicate words only,
and not to objects, and especially not to proper names, such as "Ivan,"
"Moscow," etc. ). In decoding, all the coordinates of the initial vectors
Should be zero.

The distances pr (a, b) and px«(a, b) are then determined for these words,
and compared with each other. If the consistency criterion is satisfied,
the routine is terminated. Otherwise, the system of semantic categories
is altered.

Two cases may arise:

1. For some word pairs from (a,b), pr(a, b) « o«(a, b). In this case, the
coordinates which are different for numerous word pairs from (a,b) are
eliminated from the list of coordinates (the column vectors). This lowers
the dimension of the semantic vectors because the column vectors dif-
ferentiating between words which are close in a given text are omitted.

2. For some pairs from (a, b}, er(a, b) px(a, b). In this case, new
semantic coordinates are introduced in the following way: we search
for a word c such that

d — pz (a, c) — pr (c, b) = max. (4.7)

This word c corresponds to a new semantic category (coordinate) which
takes the value 0 for the words d satisfying the inequality

er (c d) « er (c, 5) 5, (4.8)

and the value 1 for all other words (for the word 6 this coordinate is 0,
and for a, it is 1).

In other words, if the word c is "table," introduction of a new semantic
category corresponds to classification of words according to the presence
or the absence of the property of "tableness."

We *hus obtain a new system of coordinates, different from the original
System. It is checked using the consistency criterion and, depending
on the results, we pass on to a new coordinate system or terminate the
routine.

194

$10. CLOSENESS-IDENTIFYING ALGORITHMS

Algorithm determining the graph of syntactic connections
of words in a sentence

We have mentioned earlier that the visual distance between the elements
of a message does not correspond to the intuitive "closeness of meaning."
In other words, the intelligibility of a message (i.e., the ability to predict
the content of the inaccessible part after examination of the accessible
part), although relatively low if the visually observed closeness relations
are used, may be markedly enhanced if we pass to optimal relations.

For example, if the message is a linear succession of signals, the
successive signals may be regarded as maximally close; however, if the
message is a scan of a two-dimensional image, signals separated by a
single line length are also maximally close.

Therefore, if the particular signal represents a dark detail in a
pattern, a similar black-dot signal can be expected not only in
immediate adjacency to the first signal, but also after a certain interval.

Words of an "ordinary" language related in meaning do not necessarily
occur one next to the other, either. The arcs in the sentence in Figure 63
[from Walt Whitman's Song of Myself] connect words which are more
"closely related" than the unconnected words.

the flag of my disposition, out of hopctul green stuff woven

FIGURE 63. Syntactic linkage of words in a sentence "... the flag of
my disposition, out of hopeful green stuff woven" (Walt Whitman,
Song of Mysclf).

Pairs of linked words are meaningful, albeit sometimes "half-baked,"
e.g., "flag of," "of hopeful," "hopeful stuff," whereas other word pairs are
meaningless, such as "green woven," "disposition stuff," etc. There is
a very close syntactic and semantic linkage between the two extreme words
of the sentence, "flag woven."

In some texts (e. E., in Latin poetry), the great visual distance between
words related in meaning gives the impression of an intentional jumble
(Figure 64).

In nova fert animum mutatas dicere formas corpora

FIGURE 64. Syntactic connections between words in a Latin sentence

195

EXTRATERRESTRIAL CIVILIZA TIONS

Retaining the original word order, the sentence can be translated as
follows: "In new attracts the soul changed to tell shapes of bodies,"

i.e., the soul is drawn to tell how bodies change into new bodies. Know-
ledge of the true relations is essential for detecting higher level units,
since these units consist of "close" units of a lower level.

In describing the basic version of the algorithm below, we shall assume
that both the lower-level units — words — and the higher-level units —
Sentences — are known.

This algorithm should establish a "true" semantic closeness between
words in a sentence, or more precisely, in a simple sentence, i.e.,

a sentence which does not contain other sentences.

Our problem will be solved if we identify pairs of words in a simple
Sentence which are directly related in meaning.

A pair of words directly related in meaning can be described as a
segment of a line whose ends correspond to the particular words or to
some symbols replacing the words. Then the entire set of words ina
sentence which are directly related in meaning will be represented by a
drawing, or what we calla graph, of the general form shown in Figure 65.

aS people
cloud from i
22 slowly N decided
black the west meeting
FIGURE 65. Pairs of words con- FIGURE 66. A graph with inde-
nected by straight segment: are termipate meaning.
meaningful.

The shape of a graph characteristic of a simple message can be
predicted in advance. The prevailing opinion is that the graph of a simple
message is a "tree, " or in other words a graph without cycles. A graph
is said to be connected if, by retracing its segments (sides) we can pass
from any vertex to any other vertex; a connected graph is called a tree
if there is only one such path between any two vertices.

This is a reasonable assumption, which indicates that all the words in
a sentence are connected in meaning, though possibly not directly.
Moreover, these connections are unambiguous, whereas the graph shown
in Figure 66 is ambiguous; there are two possible interpretations of the
Sentence, "decided meeting people" or "people decided meeting."

The basic assumption used in constructing a quality function is that
there is a strong predictive link between words which are directly
connected in a sentence. However, to apply this principle in practice,
we have to set up a certain classification of words. First, words occur
fairly infrequently, and pairs of words are even more infrequent. The
introduction of classes enables us to group numerous words into a single
category, representing the individual words by the appropriate class
symbol. The class symbol will occur frequently in the text. Second, the
alphabet of words is very extensive. Unless it is compressed, no computer
experiments will be able to test the algorithms.

196

IV. MESSAGE DECODING

The classes must not be defined at random: "good" classes should be
introduced, with the aid of a special decoding algorithm. We cannot
propose any particular algorithm of this kind at present, but we have a
rough idea of what they should look like.

The algorithm probably will be based on the so-called "grammatic"
classes of words. In our specimen calculations for a Russian-language
text, we used the following selection of classes: cases of nouns and ad-
jectives, finite verbs, adverbs, verbal adverbs, infinitives, prepositions,
conjunctions, particles.

Each word is then assigned the symbol of its class, and the text is -
decoded using these symbols.

Consider how words (or, more precisely, classes of words) predict
one another in a sentence. Let us compute the conditional probability
of a word of class kj occurring in a sentence if it contains a word of
class ki. This conditional probability can be computed from the equality

| P Gu ky) P (kn, ki) — e Ry)
pk) Vu ~ P) "7 e)

Here p(k; k) are the probabilities of the joint occurrence of words of the
classes k; and kj within a simple sentence, p(k;)is the probability of oc-
currence of the symbol k;in a simple sentence. P and ọ denote relative
and absolute frequencies, respectively.

To estimate the "mutual predictability" of the symbols b; and&,, we
form the average of p(k;/A;) and p(Aj/kji), denoting it P (kn &;).

A partial predicate system for a given text is defined as the square table
of numbers p(£Ri, kj).

If a certain sentence tree is given, we can assign a weight p(k,(%u), &,(.))
to every side of the graph joining the words A, and A,. This weight is a
function of the classes k; and k; and the words A, and A, at the two end-
points of the segment.

It is assumed that words directly connected in a graph are characterized
by a high mutual predictability. We may therefore use as the quality function
of the graph the sum of the numbers p(k;(Au), &j(Av))for all the sides of the
graph.

The set of permissible solutions in our case is thus the set of all the
possible trees constructed from the symbols of the given simple sentence,
and the quality function is

D- x 2 Ò (ki (Ay), By (Ay) ),

where Au, A, are the symbols joined by the particular side of the tree.
The best solution maximizes the function D.

A similar problem has been tackled in mathematical economics to find
the shortest interurban telephone network. There are two versions of the
algorithms extremizing the function D. We will describe here the simpler
of the two, published in /6/. This method, like most other algorithms
of the so-called discrete analysis, is not exact, but it is quite acceptable
in practice. In principle, a rigorous and exact algorithm can be devised,
which would examine all the possible trees, calculate the corresponding
values of D, and choose the tree which maximizes D. For long sentences,
however, this method is too cumbersome to be practicable.

197

EXTRATERRESTRIAL CIVILIZATIONS

The set of word symbols of a given sentence is partitioned into two parts,
accessible and inaccessible. Initially, the accessible set is empty.
The first step is to choose two vertices A, and Az such that

P (k (A), k (àa) ) Z A (k (Ay), k(3))

where A; and A; are two words of the text which do not coincide with à, and
A», respectively. The vertices A, and À; are accessible and are joined
by a side of the graph.

If there are both accessible and inaccessible vertices, we search for
an inaccessible vertex A,such that

p Uu (àu), k (AQ ) > p (k (ào), k (Aw) )

where À,is some accessible vertex, A, is any inaccessible vertex other
than à» and A, is any accessible vertex. The vertices are joined by a side
of the graph, and A,is added to the list of accessible vertices. If there
are no inaccessible vertices, the routine is terminated.

This algorithm has a number of more interesting versions. We can
indicate the order of best reading of the sentence, i.e., the examination
procedure which ensures the fastest decoding of meaning. The "main
ideas," or more precisely, the main words are the first to be grasped
in this way. We can thus fix a certain order of preference or a subordination
relation for the words in the sentence.

To this end, it suffices to indicate directions on the sides of the graph
connecting the different words. These directions are determined from the
following considerations: a subordinate word more strongly predicts its
principal, is "in a greater need of the principal," than the other way
round. Therefore, if p(kE(A;))/k(Aj)) > p(R(A;)/R(Ai)) the arrow should point
from A; to A.

If we read the sentence in the reverse direction, the sense of un-
intelligibility and confusion will persist for a longer time, since the
already examined subordinate words strongly predict their principals,
which still remain inaccessible.

Anyone who has ever tried to learn German and Latin is familiar
with this curious feeling!

An algorithm identifying "types of syntactic relationship"
of words

The algorithm described above has been checked manually for small
examples only. However, before proceeding with serious experiments,
we should analyze some of the errors which this algorithm introduces.

Apparently, most of the errors are associated with the extreme
imprecision of the description of the different words: after all, even
words of the same grammatic class are markedly different from one
another. For example, the sentence "structure of hydrogen atom"

[in Russian — stroenie atoma vodoroda] is coded as n,g,,2,.^ If

* Nominative case of a noun, genitive case of a noun, genitive case of a noun.

198

IV. MESSAGE DECODING

na B(ns Ea) > Ble» ga) (as is to be expected), the
algorithm will produce the tree shown in
Ea [A ny £j— 8, Figure 67a, whereas the correct graph of
2 b this sentence is b in the same figure.
It is readily seen that connectedness of
FIGURE 67. Incorrect (a) and correct words is sometimes independent of their
(b) graphs of the Russian sentence grammatic form; for example, the graph
"structure of hydrogen atom" — for the part of a sentence which reads "by

stroenie atoma vodoroda. structure of hydrogen atom" [stroeniem atoma

vodoroda] in Russian is evidently independent
of the case of the first word and therefore coincides with the graph of the
original sentence "structure of hydrogen atom" [stroenie atoma vodoroda].

The algorithm described below is intended for avoiding these errors;
it is also of independent interest.

Suppose that words are described using two classifications, and not
one as before. One of these classifications is a grammatic classification,
and it will be identified with the classification previously described.
Another classification is a lexic classification. We assume that the classes
of the lexic classification contain words with a common base. Each word
is thus described by two symbols: the symbol of its grammatic class
and the symbol of the lexic class.

A "partial predictive system" for this description of a dictionary can be
presented in the form of a table where the row g; contains numbers cha-
racterizing the grammatic class gi, and the column under /; contains
numbers characterizing the lexic class /;,. The entry in the square i, j
in the table is the conditional probability that a sentence containing a
word of class i will also contain a word of class j. The numerical
entries are calculated by analyzing a certain text

fi. £m PPP de

Once the table has been computed, we can calculate the so-called
connecting function o, for any two words A; and ^:

o, (Ai, 4) = p (eA Mg) + pGOJALO)) +
+ pO Ms 04)) + p G (y 0 ()).

This function coincides with the conditional probability, but it has the
advantage of enabling us to described words in terms of two different
classifications. `

We have already noted that the lexic grammatic classes coincide with
the classifications of syllabics into vocalic and consonantal. A given pair
of classifications therefore can be identified in principle by an algorithm
similar to that mentioned on p. 191.

199

EXTRATERRESTRIAL CIVILIZATIONS

As in the first version of the algorithm, the table of conditional
probabilities is replaced by the table of mutual predictabilities, i.e.,

numbers of the form p(k,, k)-ip (ki/k;) + p (kilki; the connection function
oz calculated using the table ||5(k;, &;)| takes the form

95 (A, Ay) = P (Eli), 2) + A lel), 10) +
+ BOO, g(a) + BU 04) 1(.)).

Consider a sentence where every word has been coded with a pair of
symbols of this kind. We can then construct a tree of this sentence as-
signing the numbers 0; (A; àj) to the sides of the graph and applying the
procedure described above.

Errors of the kind encountered in the sentence "structure of hydrogen
atom" are corrected because the connecting function allows for the frequent
joint occurrence of the two words "hydrogen atom" in a single sentence,
irrespective of the cases (this fact is expressed by the high value of the
term (gA) (Aj)). The sides of the graph are dírected using the same
considerations as before.

This algorithm yields qualitatively new information about sentences.

We can establish four types of connection, depending on which of the

terms makes the largest contribution to the value of the connecting function
oz, P (A) /9(As)), PEAD OAD), p (10/1 (Ai) or p (L(:)/2()).. Their interpretation
is readily understood if we remember the concept of "subordinate clauses"
taught in high school. Knowledge ofthe various connections enhances the
intelligibility of the text, since it provides the possibility of direct pre-
diction of words. Let us consider the different types of connections.

Il.g-g type. A form of connection whereby the grammatic class
of a word à; strongly predicts the grammatic class of a word à; an effect
generally called consistency.

Tl i +g type. The lexic class of a word 4; strongly predicts the
grammatic class of a word 4;. In traditional grammar this is known as
"government."

IH. | —] type. The lexic class strongly predicts the lexic class of
another word. The adjoining effect of traditional grammar. According
to the traditional approach, this type of connection is characteristic of
words which do not change in declension, e.g., "very early" (it is not
clear to what extent this definition is consistent with the traditional
definition of the concept; the same holds true for the other types,
however).

IV. g —1 type. Without analogy in traditional grammar. May not
occur in reality altogether. No obvious examples known.

The simplest algorithm of literal machine translation

We will now consider one last generalization of the sentence-graph
algorithm, which although of unquestionable importance, requires a great
deal of preliminary information about the text.

We will show that this version of the algorithm may be useful for
machine translation (possibly after some modifications). We mean here the
Simplest and most elementary form of machine translation, the so-called
literal translation, when the translation process assigns a certain word to
every significant word of the original.

200

IV. MESSAGE DECODING

The difficulty in this translation is that the same word of the original
lends itself to different translations. If there is a dictionary which
assigns to the words of the source language all the possible translations,
the main problem of the translation algorithm is to reject all the redundant
alternatives, retaining only the most accurate and fitting translation.

The higher the number of the alternatives rejected, the more sophisticated
is the algorithm.

The complete dictionary should translate the English word "hand" into
Russian as 'xucre (human hand), "'crpeaka'(of a clock), and also give the various
declensions, etc. In this case, the choice of the best alternative would
involve choosing the correct word in a correct grammatic form. This
routine requires a system of semantic classifications, possibly similar
to that described on p. 192, and also certain "grammatic classifications."
We should be able to describe the words of the translation language using
this classification system, i.e., to every word of the translation we should
be able to assign its description vector (see p. 193). The list of symbols
of the semantic classes will constitute a certain "semantic" alphabet.

For the translation language we then can construct a square table whose
rows are identified with the symbols of classes and the entries are the
conditional probabilities p(k,/k,) (this table is computed using an extensive text
in the translation language). The connecting functions can be defined
in the Same way as before:

O(n M) e DZ p, Aka (A),

where n is the total number of classes:

n n

9404, A= X 2 BG). Ro OU).

uel vel

The function g, takes on large values for pairs of words which strongly
predict each other in the translation language.

To describe the remaining steps of the algorithm, consider a schematic
diagram of some sentence in the source language and the alternative
translations performed with a bilingual dictionary.

First word Second word Third word
+ +
Translation +
alternatives + + +
bs

One word should be picked out from each column. If every word of each
column is joined by a line with every word of all other columns, we obtain
a graph which contains all the possible graphs of all the possible sentences
(Figure 68).

From this graph, however, we should select only one tree, with one
vertex in each column. This graph can be found in the usual way (with
slight modifications); the only difference is that we are not looking for a
graph connecting all the vertices: in each column, all the vertices but

201

EXTRATERRESTRIAL CIVILIZA TIONS

one should be isolated, as in Figure 69. The isolated vertices are the
rejected alternatives.

+ + +
+

+ +

+
+ +

t *
FIGURE 68. A graph containing all FIGURE 69. A probable end
the possible graphs of the translation product of a graph-selection
sentence. algorithm,

$11. MATCHING ALGORITHMS (END)

Morpheme -identifying algorithms

As we have noted before, morphemes are the smallest meaningful
units of the human language, i.e., the set of morphemes is the alphabet
of elementary signals for semantic levels.

The word "unloader," for example, is divided into morphemes as
"un+ load+ er." Certain morpheme sequences form words. Words in
various orthographies, although by no means in all, are separated
by special signs — blanks — from one another. Morphemes, as a rule,
are not separated from one another. "Therefore, in automatic text
anelysis, the word boundaries are assumed to be known, whereas the
boundaries between morphemes are sought by a special algorithm.

However, we have no right to expect a message from space to follow
the pattern of terrestrial writing. We shall therefore have to aim at
the most difficult case, namely when neither the word boundaries nor
morpheme boundaries are defined to start with. In such a case, we should
start with an identification of code groups. The main difficulty is that
the code groups forming morphemes are of varying length. Therefore,
the set of permissible solution is the set of all possible partitions of the
text, the number of which is (V — 1! There is one further significant
difference: the number of morphemes in human languages is much
greater than the number of letters which represent sounds. This is not
a fundamental difference, but a simpler procedure would probably apply
to the decoding of an inhomogeneous code representing letters.

The algorithm described in what follows uses a quality function which
is different from V and a different recognition procedure. Because of
these differences, the algorithm will identify higher level groups of letters,
and not only morphemes. Groups of morphemes are of interest if they

202

IV. MESSAGE DECODING

constitute words or groups of words. The algorithm described below
identifies not only morphemes, but also words (at least some) and
certain groups of words.

There is of course a possibility that certain combinations of morphemes
will remain "unresolved"; a special algorithm will be required to partition
these combinations into the constituent morphemes.

Because of the change in the purpose of the algorithm, the set of
permissible solutions is different. A correct partition of the text by
round parentheses is defined as the original text with symbols of two
kinds — right and left parentheses — interposed. The right and the left
parentheses are placed in accordance with the following rules:

1) a correct group of zero order is a group of letters of the original
text enclosed in a left parenthesis on the left and in a right parenthesis
on the right;

2) a correct group of i-th order is a group of correct groups of
(i—1)-the order.

A correct partition of a text by round parentheses is obtained if the
parentheses interposed in the original text convert it into a regular group.
It follows from this definition that correct groups of the same order
may "link up," i.e., the beginning of one regular group may coincide with
the end of another group, if the one is not contained completely in the other.

This concept establishes the complete separability of morphemes, which
are never intertwined into a jumble in a sentence.

Simple groups correspond to morphemes; this, however, does not
take into consideration the possible inclusion of morphemes in one another
(man — men) and other exceptional effects. It is impossible to allow for
these effects without markedly complicating the algorithm.

The quality function will be derived from the following considerations.
While analyzing the previous algorithm, we noted that for code groups of
equallengths, the difference between the groups of correct and incorrect
solutions is that the parts of correct groups are "connected" more closely
than the parts of incorrect groups. This concept constitutes the basis
of the new quality function.

A group is said to consist of "strongly connected'' parts if the
appearance of a certain part of the group strongly predicts the appearance
of the remaining part. For example, ifa certain text contains the group
"envelo," then it is almost certain to be followed by "pe." This propo-
sition holds with varying degree S of likelihood for other partitions of the
word "envelope." The predictability is naturally calculated using the
expression for the conditional probability: given an i-th partition of a
word A into right and left parts, we can write the word in the form A-LiRi
(where L; is the left part of the i-th partition, R; is the right part);
the predictability of the right part by the left part of the word can be

expressed by the fraction BEARD The predictability of the left part by

the right part is similarly expressed by the fraction SOM.

On the whole, the quality of a group is determined as follows:

1. 'The group is divided into two parts in all possible manners.

2. For each partitition, the predictability of one of the parts by the
other is determined.

203

EXTRATERRESTRIAL CIVILIZATIONS

3. The mean of all the predictabilities corresponding to the various

partitions is calculated.
Since in a chain of length d we may introduce d— 1 partitions, the total

number of predictabilities (i. e., fractions of the form SER) ang SLIR) )
9 (Li) 9? (Ri)

is d — 1. The quality of a group or its "stability" is therefore expressed
by the equation

h 7 gg e(L) T e(R)

d~i
ad (LR y LRA),

i=l

This relation is naturally inapplicable to evaluating single -letter
groups: the stability of these groups is taken equal to zero. The stability
of groups which do not occur in the text or which occur only once is also
set equal to zero; the latter choice is explained by the fact that any group
of the text including an incorrect one, occurs at least once.

The quality of the partition Y(R)is calculated as the sum of stabilities
of all the correct groups entering the particular partition.

It would naturally be impossible for us to examine all the correct
partitions of a text by round parentheses and to calculate for each partition
the quality function Y(R). We will therefore propose a routine which, as
always, ensures a fairly high value of Y(R).

We will require a frequency dictionary of all the groups occurring in the
text, i.e., a list of these groups indicating the number of occurrences
of each group in the text.

The simplest method for compiling such a dictionary is the following:
first we draw up lists of all the possible groups of length 1,2,..., etc.,
and then examine the text and count the number of times the particular
group occurs. This method, however, is unsuitable for fairly long groups.

The compilation of the frequency dictionary is substantially simplified
since we are not concerned about groups which occur less than twice (their
stabilities are zero). We can therefore use the following routine: draw
up a list of letters and determine their frequencies; letters occurring
less than twice are dropped from the list; the frequencies of the remaining
letters are included in the finallist. The resulting fragment of a frequency
dictionary is called a first-order fragment.

If a fragment of order i —1 has been compiled, a fragment of order i is
constructed in the following way: for every groupAfrom the fragment of
order i —1, we construct all the possible groups of the form Aa; where a,— 1
is some letter of the alphabet; then determine the frequencies of these
groups and omit all chains which occurred less than twice. The list of
remaining groups and their frequencies constitutes a fragment of order i.

The successive construction of fragments is terminated when a fragment
of the next higher order proves to be empty. This stage is reached when
all the groups longer than any of the previously constructed groups occur
less than twice in the text.

The frequency dictionary can be markedly shortened by omitting
allthe groups which are contained in some longer group of the same
frequency. If some group A, occurs & times, any group A, contained in
Au occurs at least k times; therefore if the frequencies of A, and A,
are equal, there is no need to include the group 4, in the dictionary.

204

IV, MESSAGE DECODING

Once sufficient information is available on group frequencies, we can
find their stabilities. The computed stabilities are also included in the
frequency dictionary together with the corresponding frequencies.

The frequency dictionary of groups is then converted into a frequency
dictionary of correct groups.

The assumption which ensures the first step of the routine amounts
to the following: it is assumed that the most stable group always enters
the text correctly, i.e., it does not link up with any other correct group.

To explain this assumption, note that some groups form morphemes
or combinations of morphemes in certain parts of the text, and not in
others. For example, "un-" is a morpheme in "unloader,' but it is
not a morpheme in "bunloader.'

In the frequency dictionary, we should thus find the most stable group
‘1, and then enclose in parentheses each occurrence of this group
. in the text.

For the same reason, none of the groups linked up with any of the
inclusions of the group 4, may be a correct group. We should therefore
reduce the frequencies of all the groups in the dictionary whose inclusions
arelinkedup with the inclusions of 41; the reduction should be equal to the
number of linkages of the corresponding groups with A.

Groups whose frequencies should be reduced can be found by considering
the inclusions of 4; according to the following scheme:

k 21 1 2 k
A (aisi oe. Grass iia [+ [tiai anna - + ]

Here the group ai, ... Qiu represents the inclusion of a correct group
inthe text; i is the running number from the beginning of the text.

In the frequency dictionary, we locate the groups enclosed in square
brackets in the order of their increasing numbers; k+ 1 is the number
of the first group which has not been entered into the frequency dictionary.

The frequencies of those groups which are located in the dictionary
are reduced by 1; the square brackets are then extended to the right of
the interval between aj4;-; and aiu», and then to the right of the interval
between aj,,-; and ai4;-3, and so on, until we reach the interval between
ai and a2. Similar series of brackets are interposed to the left of the
group di ... Qip

All the inclusions of the group 4; are examined in this way. As a result
of the application of this procedure, the frequency dictionary approaches
a list of correct groups, since the number of interlinking groups in the
text diminishes. If the reduction in frequencies brings the frequency
of some group below 2, the particular group is omitted from the
dictionary.

In general, a frequency dictionary can be considered as an approximation
to a list of regular groups of finite stability. The quality of the list can be
estimated using the relation

Y’ = D Y A) Peoia).

Here A; is a certain group, q4,(A;) is the number of correct inclusions of
^; in the text.

205

IIl. RADIO COMMUNICATION WITH EXTRATERRESTRIAL CIVILIZATIONS

values of the original function, £(/4) — x(fj)). The question is, are these
functions equal at any time /, and not only at /,, i.e., are they identically
equal? The fit between the two functions is naturally improved if the
original function varies slowly between the quantization times ¢,. This
means that the function should not contain very high harmonics. According
to Kotel'nikov's theorem, the two functions are identical if the original
function x(/) does not contain components with frequencies v higher than
Af, i.e., if the band width of the Avtransmitted function is equal tothe band
width of the communication channel. Kotel'nikov's theorem is highly
Significant for the theory and technology of communication, since it
permits converting continuous functions into a train of some discrete
magnitudes for transmission. This theory maintains that a function with

a bounded spectrum Av is completely determined by its values measured

at intervals At='/,Av. In particular, a function of duration M, i.e., a
function which does not vanish only for fo«(« tg-- Af, is determined by a set
of 2\t\fdiscrete values. Thus, the definition of information derived for
discrete messages can be safely applied to continuous functions with a
bounded spectrum.

When continuous functions are transmitted by means of pulsed signals,
the main difficulty is that the function may take on any instantaneous
values, including irrational and transcendental numbers with an infinite
number of significant digits. Theoretically (in a noise-free channel),
these numbers can be transmitted with full faithfulness by PAM or another
suitable technique. In reality, however, reconstitution of the original
pulse with sufficient accuracy (or transmission of a sufficienily high
number of significant digits) in a noisy channel requires an excessively
high signal-to-noise ratio in the communication channel. "Therefore, the
next step adopted in the transmission of continuous functions calls for
quantization of the message. To quantize the message, we select from
among all the values of x(t) a set of N discrete allowed levels x, xs» ... xx,
which are distant Ax from one another (the quantization gap). All the
other values are regarded as forbidden. Only the allowed values are
transmitted. If the true instantaneous value of the function falls inside
the interval (x; xin) i.e., takes on a forbidden value, the nearest allowed
value, differing from the true value by less than half the quantization
gap, is transmitted through the channel. This operation is completely
analogous to the rounding -off of numbers; it essentially signifies that
we are transmitting the true values of the function up to a certain number
of significant digits.

The quantized values of the signal in the communication channel are
affected by random noise. The width of the quantization gap should be so
chosen that with a given probability p the noise does not exceed half the
quantization gap. Then the signal can be accurately reconstituted at the
receiving end of the channel, since in this case the signal level nearest
to the noise-distorted value is the same as that fed into the communication
channel. The probability of signal reconstitution error is equal to the given
value p. The reconstituted signal can be again sent through the communica-
tion line, and this procedure may berepeated severaltimes, without affecting

206

IV. MESSAGE DECODING
Letter-identifying algorithm

We mentioned on p. 148 that linguistic phenomena are best characterized
by a selection of distinctive features, but the question of how to choose the
set of distinctive features still remains open. Consider the following
situation: there exists a set of "elementary features." These elementary
features are then combined in an optimum manner according to certain
rules to produce compound features. A certain criterion (e. g., a quality
function) is required to assess the quality of these features.

The simplest part of this program is apparently the choice of the set
of elementary features. The particular choice, it seems, would be
largely immaterial.

Let us consider the elementary features for messages received
visually. This category includes written texts, pictures, and the entire
visible universe. It is clear that the elementary features should constitute
the simplest elements of sensation associated with the smallest differences
noticeable to the eye (detectable with our detector).

If every color is regarded as a mixture of the three basic colors — red,
blue, and yellow — the visible universe can be represented as a six-
dimensional space, with each point described by the six coordinates
QrERS Opt Eg, Ay rly, Q8, Qu'En Ant EnWhere er ey, ey are the minimum
increments in the intensities of the three colors, e,. ¢,, e, are the minimum
detectable widths, lengths, and heights, expressed, say, in angular units.
The symbols containing the letter e are unit vectors or "elementary
features"; symbols containing the letter a correspond to the strength or
the magnitude of the feature in the particular object.

Similarly, acoustic impressions can be described in terms of elementary
features associated with minimum detectable acoustic differences, but this
approach is not absolutely essential: sounds can bedecoded and presented in
chart form (e.g., an oscillogram).

In general, we receive information about the outside world in the form
of what is known as "sensation"; signals which are not detected directly
by our sensory organs are first converted by "physical instruments."

On the other hand, the transition from microevents (points in the
space of elementary features) to more complex units, e.g., letters in
the usual sense, is in general a highly complex problem.

We will now consider, in a highly approximate form, the problem
of letter identification, assuming that we have in our possession an
instrument which is capable of distinguishing between black and white
squares (on a sheet of paper covered with a fine grid) and identifying the
position of the squares.

This problem belongs to the domain of pattern recognition. Extensive
literature is currently available on the subject.

Note, however, that the great majority of sources treat the problem
from the aspect of learning to recognize objects. Procedures based on
this principle are constructed as follows: consider a certain set of objects
presented to the computer; the computer has a certain number of responses.
The computer should assign one of its responses to every object (i.e., in
practice, it should generate a certain classification of the objects). If the
machine "errs," the teacher — a human operator — informs the computer

207

EXTRATERRESTRIAL CIVILIZATIONS

of its error and "penalizes" it; otherwise, the computer is "rewarded."
When the learning stage is completed, the computer is capable of recog-
nizing the objects independently.

The learning approach is naturally ruled out in decoding problems.
Moreover, simple classification of objects is not enough in our problems:
the boundaries between different letters must be indicated.

If we ignore for the moment the problem of combining the "images"
of the same letter into classes, the problem of letter identification
becomes similar to the problem of identification of simple morphemes.
We may assume with fair accuracy that letters occupying non-overlapping
parts of the surface and the combinations of dots filling these areas are
Stable in a certain sense.

Letters are not necessarily similar to meaningful images. Letters
need not be smooth or connected: they may represent a random combination
of points. It suffices that all the inclusions of one letter correspond to
roughly the same combination of points.

A distinctive feature of the problem of letter identification compared
to the problem of morpheme identification is that no two identical in-
clusions of a single letter exist: there are only similar inclusions of
varying likeness. The basis for this likeness is extremely difficult
to detect: sometimes a slight change in the outline coverts one letter
into a different letter (e. g., the Russian letters H and M, and IU),
whereas much more radical morphological changes leave the letter
unchanged (oe and c, 0 and d). The size of letters, their inclination,
and the degree of stretching generally do not matter, although the cursive
e and l differ in size only, the Cyrillic E and IW are in fact the same
pattern rotated clockwise through 90°, p and b have the same shape,
rotated in a plane and reflected.

In certain writing systems, e.g., shorthand, even less conspicuous
features are decisive for letter pronunciation, e.g., elevation above the
line level, thickness of strokes, etc.

The most effective algorithm should apparently contain a set of rules
for building up elementary features into really distinctive features of
letters. Some of these features are probably recognizable only in the
presence of other letters used for comparison (e. g., elevation above line
level, inclination, size).

Letter identification without reference to nearby letters is therefore
an impossible undertaking. We can only hope that in most cases it
suffices to examine a very small neighborhood of the letter being analyzed.

The algorithm described below does not pursue any serious aims.
Nevertheless, it is of a ceriain interest because it uses a very limited
amount of initial information (in particular, letter sizes need not be
considered) and leads to a definition of frequency which is far from self-
evident.

Let us first define the so-called residual similarity. Let an element a
be located ın some area K,of the text. Suppose that this element is de-
scribed by a function6(x,y), where x and y are the rectangular coordinates
of the points in that area, and 6(x,y) takes on two values only: 1 for a black
dot, and 0 for a white dot.

Let an element 6 be located in another area K, and suppose that this
element is described by a function ð; (xı, yi), where x,y, are the coordinates
of the points of the second area in its own system of axes. "These are also
rectangular coordinates.

208

IV. MESSAGE DECODING

The coordinate axes of the two areas are translated until their origins
coincide and the respective axes are parallel. The residual dissimilarity

+ is defined by the expression

1 1
s^ltikngi f fa (x1, y) — 6c, y) Pdx dy.
dn Ko

The symbol S stands for "residual similarity." The symbol KNK:
indicates intersection of the corresponding areas.

The following transformations are allowed for the coordinates of the
area Ky: 1) paralleltranslation, 2) similarity transformation, 3) contraction
along each of the axes, 4) change of angle between the axes, 5) rotation in
a plane, 6) mirror reflection.

The function 6, and the coordinates xi, yı are transformed so that the
values of the new function in the new coordinates coincide with the values
of the old function in the old coordinates, i.e., 3

ô! (xl, yt) = 8, (Xv Ji)

where 8| (xl, yi) stands for the new function in the new coordinates. Under
these conditions, the change in the function will be completely determined
if we specify the transformation of coordinates. Ifa new function is
expressed in the old coordinates and the residual similarity with 6(4, y)

is determined, we will find that it has changed compared to the residual
similarity before the transformation. 'Thus, each transformation of the
original coordinates can be assigned a certain value of the residual
Similarity.

Let the six transformations define six axes in the transformation
Space. Along each axis we lay off the "values" that each transformation
may take (it is assumed that a single-valued quality function has been
defined for each transformation). The points of this space, defined by
combinations of the values of the quality functions, will be called
permissible points. At every permissible point, two functions are
defined: a scalar function — the residual similarity, and a vector function —
the gradient indicating the direction of fastest growth of the residual
Similarity. Moving along the gradient, we can find a point at which the
residual similarity reaches its maximum value for the two areas and the
given elements. This maximum value will be used as the true similarity
of the two elements a and b. We will use the same symbol S as before,
and in the following S is to be interpreted in this sense.

Consider an arbitrary area K of the text, to be used as a reference
standard. The contour enclosing this area is translated by the smallest
possible steps in the vertical and horizontal direction. For every
position of the contour, we determine the similarity of the corresponding
area to the original area. S in this case is a function of the position of
the contour. The number of maxima of this function is adopted as the
frequency of the element contained inthe initial text area. Consider a text
area made up of two squares K; and K», with a common side. 'The absolute
frequencies of the first and the second square and the frequency of the
rectangle made up of the two squares will be designated (Ki), p(K2), and
(Ki, K2).

209

EXTRATERRESTRIAL CIVILIZATIONS

The predictability of K: from Kj, or p(K2/K,), is defined as the ratio

on , and the predictability of K, from K,is correspondingly defined as

oom, or p(K,/K,). The average of these two predictabilities is defined
2

as the mutual predictability of K, and Ky, or p(Ki, Ko).

We now start increasing the size of K, and Kz, measuring the predictabi -
lity after every small increase (the straight line accommodating the
boundary and the position of the center of the boundary remain unchanged).
We thus present the mutual predictability as a function of the size of the
Squares. A plot of this function is shown in Figure 70, where D is the

size of the rectangle.

PUK Le)
The u--

! Lm t

[

>Z =

FIGURE 70. A probable plot of the FIGURE 71. A probable plor
function p (Ky, Ka) vs. the size of the function C vs. bound-
of the squares K, and K3. ary position.

Indeed, all the mutual predictabilities are less than unity: after all,
the symbols contained in pairs of squares are more individualized and
richer in detail than symbols enclosed in separate squares. Elements
showing maximum to symbols enclosed in separate squares are therefore
more frequent than symbols in pairs of squares.

If the squares are small, the mutual predictabilities are close to unity,
since the elements in separate squares and in pairs of squares are still
similar to one another, and their similarity functions have maximum at
approximately the same points of the text.

As the size of the squares increases, the mutual predictabilities
at first decrease, and then increase reaching unity, since the elements
in large separate squares, and likewise the elements in a pair of squares
(a rectangle) occur once only.

Let us find the minimum value b of the variable D such that for any
choice of the boundary between the Squares K; and Kz in the text (with
squares of the size b), the mutual predictability p(K:, K2) is unity. The
centrality of the boundary of the squares K, and K3 is defined as

b
C= [Pik K,)(D)dD. The centrality characterizes the mutual predictability
0

of the squares, irrespective of their size, and is thus a fundamental cha-
racteristic of the boundary location.

The basic hypothesis of our algorithm is the assumption that centrality
is minimized (compared to other near positions of the boundary) when the
boundary passes between the letters and is aligned along the true "physical"
boundary of the letters. Then we start moving the boundary in some direc-
tion, measuring the values of C after each step. We obtain a plot of the

form shown in Figure 71.

: 210

IV. MESSAGE DECODING

Fixing the boundary at a point corresponding to a local minimum, we
will rotate it about its midpoint, measuring C at minimum angular inter-
vals. We select the angle a corresponding to a minimum value of C and
move the boundary infinitesimally in the given direction. In the new
position, we again select the best direction for further displacement, and
act as before. If several best directions are available, we choose the
rightmost. If we are really moving along the letter contour, we will
describe a closed curve enclosing the letter completely.

The element enclosed within the curve may be used as a reference
standard, and we then calculate all the maxima of S in the text using this
Standard.

The textual elements identified by this procedure are not quite the
letters of the alphabet. For example, the letters p and b are definitely
different, and we will therefore call this a skeleton alphabet.

Without going into details, we can outline a general procedure for
developing this skeleton into a proper alphabet. Each occurrence
of a skeletal element is assigned a vector from the transformation
Space which transforms it, say, into the first occurrence of the element
in the text. Some of the component values of this vector are stable, i.e.,
strongly predictable by the nearby elements. For instance, the letters
of the word "bed" strongly predict the variety of the symbol "b" with the
stroke directed upward, whereas the letters of the word "pet" predict
the same symbol with a downward stroke, Accidental changes of symbols
are not predicted with any stability.

$12. CONCLUSION

In conclusion, we wish to call the attention of the reader to one
remarkable aspect of decoding.

How strange to our conception may the reality hidden in extraterrestrial
messages be? Will we at all be in a position to comprehend the content of
these messages? ;

It should be stressed at the outset that there is an enormous difference
between understanding the message and comprehending what it is all about.
While intelligibility is based on the ability to predict the inaccessible parts
of the message or future events, comprehension draws upon our ability
to translate the message into the language of images corresponding to
real situations.

Not all that is intelligible is comprehensible. We cannot comprehend
the sensations of a being which responds to radio frequencies, but this
does not detract from the intelligibility of his behavior. Therefore, even
if the "other" reality is fantastically strange in our eyes, it need not be
considered unintelligible.

211

EXTRA TERRESTRIAL CIVILIZA TIONS

Bibliography

10.

11.

12.
13.
14.

15.

16.

Apresyan,Yu.D. Idei i metody sovremennoi strukturnoi lingvistiki
(Methods and Ideas of Modern Structural Linguistics). Moskva.
1966.

Hrozny ,F. Khettskie narody i yazyki (Hittite People and Languages). —

Vestnik Drevnei Istorii, Vol.2, No.3. 1938.

Dobrushin,R.L. Matematicheskie metody v lingvistike (Mathematical
Methods of Linguistics). — In: "Matematicheskoe Prosveshchenie,"
Vol.6, Moskva. 1961.

Kaplan,S.A. Elementarnaya radioastronomiya (Elements of Radio
Astronomy). — "Nauka." 1966.

Pratt, F. Secret and Urgent. — N.G. 1942.

Prim,R.K. Shortest Communicating Networks. — Kiberneticheskii
sbornik, Vol.2. [Russian translation. 1961.]

Romanov,V.P. Integral'nye metody opoznavaniya. Chitayushchie
ustroistva (Integral Recognition Methods. Reading Devices). —
Moskva. 1962.

Sukhotin, B.V. Algoritmy lingvisticheskoi deshifrovki (Algorithms
of Linguistic Decoding). — Problemy strukturnoi lingvistiki.
Moskva. 1963.

Sukhotin,B.V. Eksperimental'noe vydelenie klassov bukv s
pomoshch'yu elektronnoi vychislitel'noi mashiny (Experimental
Identification of Groups of Letters with a Computer). — Problemy
strukturnoi lingvistiki. Moskva. 1962.

Sukhotin, B. V. Algoritm sravnivayushchii bukvy dvukh razlichnykh
yazykov (Algorithm Comparing Letters of Two Different Languages).—
Nauchno Tekhnicheskaya Informatsiya, Vol.2. 1966.

Sukhotin,B.V. Issledovanie yazyka deshifrovochnymi metodami
(Language Studies by Decoding Techniques). — Russkii Yazyk v
Shkole, Vol.6. 1966.

Harris,Z. From Phoneme to Morpheme. — Language, Vol. 28,

No.1. 1952.

Halle, M. Fonologicheskaya sistema russkogo yazyka (Phonological
System of the Russian Language). — Novoe v lingvistike. 1962.

Shannon,C.E. and W.Weaver. The Mathematical Theory of
Communication.— Urbana. Univ. of Illinois Press. 1959.

Shreider,Yu.A. Mashinnyi perevod na osnove smyslovogo kodiro-
vaniya tekstov (Machine Translation Based on Semantic Coding
of Texts). — Nauchno-tekhnicheskaya Informatsiya, Vol.1. 1963.

Yaglom,I.M. and A. M. Yaglom. Veroyatnost' i informatsiya
(Probability and Information) .— Fizmatgiz. 1960.

212

Chaptev V

RATES OF DEVELOPMENT OF CIVILIZATIONS AND
THEIR FORECASTING

$1. THE IMPORTANCE OF THE PROBLEM OF
RATES OF DEVELOPMENT

The existing scientific notions on the possible rates of development of
life and civilization in other planetary systems may greatly influence the
attempts to establish interstellar communication.

These scientific notions directly determine the estimates of the abundance
of life and intelligence in the Universe and the number of civilizations which
are technologically more advanced than our terrestrial civilization and, in
particular, are capable of sending radio messages over thousands, millions,
and billions of light years.

The first question to consider is to what extent the historical rates of
development of the Earth civilization (and the rates of its biological pre-
history) are typical of civilizations in general. For example, were we to
accept the suggestion that the Earth conditions stimulated the biological
evolution to a greater extent than the conditions prevailing on other
potentially life-sustaining planets, we would be led to regard the Earth
civilization as one of the most advanced in our part of the Galaxy. Some
evidence supporting this idea will be given in the following. In addition to
the evolutionary factors, we also have to consider the exact time at which
life originated and the extent to which a given set of conditions fixes the
evolution rates (in principle, for identical initial conditions and environ-
mental factors, quantum fluctuations may alter the time of appearance of
significant mutations and thus change the timetable and the general course
of evolution).

Prediction of the future rate of development of the Earth civilization is
also important for establishing radio communication with distant civiliza-
tions. Different forecasts of our future ability to transmit and receive
messages may lead to different approaches to the entire aspect of the
allowed expenditure of capital and effort in this field. These approaches
may be guided by the ideas of the internal determinism of the rates and
directions of development (i.e., to what extent these growth rates will
increase following the reception of signals from more advanced civiliza-
tions) and by the importance we attach to maintaining consistently high rates
of our scientific-technical progress.

Scientists and science-fiction writers generally assume a great variety
of life forms and civilizations on different planets, but tacitly imply that the

213

EXTRATERRESTRIAL CIVILIZATIONS

variance in the rates of evolution is not pronounced. Thus, Macgovan and
Ordway /1/ give three different distribution curves for the length of time
between the inception of life and the appearance of civilization.

In each of these distribution curves, the rms deviation is 10—20% of
the mean duration. Therefore, if we take 3 to 4 billion years for the mean
duration, the evolution of intelligent life on any planet will take no less
than 1 billion and no more than 6 billion years. This approach a priori
deprives the Universe of some of its inherent richness and variety.

Even on the Earth, the rates of evolution vary between considerable
limits in different zoogeographical zones: small isolated zones (Australia,
Madagascar) are still inhabited by archaic life forms. Earth analogies
suggest significantly slower rates of evolution on planets of the size of Mars
or the Moon or on planets with small dry-land area. By the same token,
the evolution is faster on larger planets, since all other conditions being
equal, the population there includes a greater number of individuals.

The rate of evolution should drop on a planet where the climatic condi-
tions are steadily uniform (no glacial periods) or, conversely, if the
climatic eras alternate with excessive frequency. If the equatorial plane of
the equator is not inclined to the orbital plane, the tasks of the evolution
are made much easier: many of the survival problems are eliminated and
the evolution may become more sluggish or even stop altogether. At the
other extreme, if the equatorial plane is perpendicular to the orbital
plane, numerous life forms will have to migrate over large distance
annually, so that all boundaries between individual zoogeographical zones
will be obliterated and the probability of branching in the course of
evolution will be reduced. The number of large satellites of a planet
determines the pattern of nocturnal illumination and thus influences the
behavioral flexibility of nocturnal animals. This reasoning applies when
the organic world of a planet is not inherently different from that of the
Earth, especially with regard to ecology, hereditary mechanisms, and
variability of species. It nevertheless seems that the rates of evolution
of life forms and social structures — civilizations — may vary between wide
limits. Unfortunately, we still have no reliable information on the rates
of evolution and development of extraterrestrial civilizations /2/. The
analysis that follows will be based entirely on the rates of development
of our terrestrial civilization.

$2. THE ASPECTS OF DEVELOPMENT
OF CIVILIZATIONS

Although such concepts as acquisition, processing, storage and trans-
mission of information are useful in describing the development of a
society as a whole, they are insufficient for describing any particular
stage of this development. The rates of development of a civilization have
to be treated in terms of sociopolitical and economic development, evolution
of language and art, development of science and technology, the role of
religion, etc. We cannot maintain, however, that these aspects of
civilization as we understand them now will apply indefinitely to describe

214

V. RATES OF DEVELOPMENT OF CIVII IZATIONS

the progress of the Earth civilization. Moreover, there is no a priori
justification for extending these concepts to other civilizations, thus
constraining them to follow approximately the same evolutionary course.

Can we be certain, for example, that the stage of religious awareness is
equally prominent in all civilizations? On the one hand, there are indica-
tions that the development of religion is associated with local terrestrial
factors /3/, but on the other hand, even elephants are endowed with
religious sentiments and prayers /4/; if this is indeed so, religion is a
universal phenomenon. A study of the history of religion for purposes of
the general theory of rates of development of civilizations is of particular
interest because religion (especially at the later stages of its evolution)
is a clear example of a retarding force slowing down the growth of
civilization.

Another interesting factor to consider are the relative rates of develop-
ment of art and science. The discussion revolving around the topic of
"physicists and poets" which figured prominently in the Soviet press at the
end of the 1950s helped to formulate some questions, without answering
them. Feinberg /5/ noted that the rates of growth of science have long
Since overtakenthe rates of growth of thearts and humanities, and the current
trend in all probability can be extrapolated into the future. A point which
is not so clear concerns the relative significance and value of sciences and
arts in the life of a society and its individual members. Our understanding
of the laws governing the contemporary evolution of art is still far too
fragmentary to be applied in quantitative reasoning.

The development of language and other means of communication between
the members of society is another important aspect contributing to a com-
plete description of the evolution of a civilization.

Language and communication

Social intercourse between the individual members of a population,
originally the individual animals in a herd, brought about the development of
Special systems of conventional symbols, in particular vocal systems,
long before the appearance of man on Earth. The main distinctive feature
differentiating human language from the "language" of animals is its
inherent flexibility, permitting introduction of new symbols whose meaning
can be explained using only the means provided by the language itself /6/.
Note that despite the acceleration in the rates of development of various
aspects of civilization, the languages have developed over the last millennia
at an approximately constant rate, retaining some 85% of the vocabulary
for 1000 years /7/. The vocabulary started growing at a somewhat faster
rate during the last centuries and decades, mainly because of the enhanced
activity and the advances in science. The vigorous growth of civilization,
however, emerges most clearly from the new systems of symbols that
have been put into routine use: road and river iraffic signs, chemical
formulae, the notation of algebra and calculus, theory of sets and mathe-
maticallogic, library classification codes and codes for the classification
of standards and patents. The slow evolution of language is counterweighed
by the rapid change in the proportion of spoken and written words, the
radical change in the place of formulae and drawings in written communica-
tions, the prominence of slides and movie films in verbal communications.

215

EXTRATERRESTRIAL CIVILIZA TIONS

The conventional language does not even try to compete with the language

of formulae and equations. As an example, let us compare the meaning

of the word hardness in the following two propositions: gypsum has

a hardness of 2, and alloys of this kind are distinguished
by their hardness.

The possibility of distinguishing between greatly dissimilar logical meanings
of the same word is provided by the context, the semantic redundancy of
our speech. Similar multivalued meanings are characteristic of many
other words, such as qualitative, height, capacity, etc. Itis
only seldom that the speaker uses a qualifying word, such as degree,
magnitude, amount, or measure, or develops à composite word
by adding, say, the morpheme —grade (German) Itseems thatthe "words"
to be used in communication between distant civilizations will not neces-
sarily be the words of contemporary languages.

What are the quantitative characteristics of communication that reflect
the level of civilization and the rate of its development? These quantitative
characteristics are not to be sought in the structure of every individual
language, but rather in the number of different symbolic systems which
remain after maximum unification and standardization (the number of non-
redundant symbolic systems, so to say). These include the statistical
characteristics of the vocabulary /8/ and primarily the total number of
terms in current use. The scientific and technical progress is even more
clearly reflected in the statistical parameters of numbers which occur in
published texts. In the course of history, the range of orders of
magnitude covered by the printed numbers increased (one-digit, two-digit,
and other numbers), and the exponential notation (with a decimal base)
is now being used with increasing frequency. The rounding off of approxi-
mate numbers is now better motivated, and it is only seldom now that
numbers are truncated at units. With the spectacular growth in orders of
magnitude, the power exponent is now often rounded off, and one may
prefer writing 109 rather than 109. These tendencies are fundamentally
simple, they can be studied without substantial material expenditure, and
they are associated with highly general and basic principles. There
seems to be no reason why these tendencies should not be generalized to
extraterrestrial civilizations.

The relative proportion of the exact (discrete) and approximate
("continuous") numbers in published texts follows a more complex evolu-
tionary pattern. Medieval texts reveal a predominance of discrete
numbers: the number of people, objects, operations; the figures of
celestial bodies and their orbits are regarded as ideal spheres and circles.
The numbers 7 and 3 are reported more often than they occur in reality.
The age of enlightenment and the development of capitalist society led to
the rejection of some superstitions, substituting measurements for counting
and philosophizing. The attention of the engineers was focused on physical
processes, primarily those involving power and energy. The discrete
numbers no longer filled the foreground. But at the end of the 18th century,
the physicists suddenly rediscovered the significance of discrete quantities:
the elementary charge was discovered, and then the microcosmos was
quantized. Soon after that, both the life sciences and technology switched
the main emphasis from energy processes to information concepts.
Morphological descriptions were often supplemented by structural charac-
teristics, and the "discrete" mathematics received a great impetus. The

216

V. RATES OF DEVELOPMENT OF CIVILIZATIONS

extraterrestrial civilizations will apparently follow a similar course of
development. i

Conversely, the great variety of the living languages on our planet and
the absence of any obvious relationship between the existing language
families can be attributed to the peculiar features of the Earth topography,
difficulties of sea voyage, etc. Ona planet with a more compact dry-land
area, fewer mountains and deserts, and a greater abundance of navigable
rivers, a single universal language may develop at the dawn of civilization
Another sufficient condition for the development of a common language would
be the availability of excellent means of transportation or communication.

'There are no reasons to believe that a civilization is incapable of
creating an artificial universal or international language. Why did this
not take place on Earth? Latin was the international language of scientists
and theologians in the Middle Ages, but it was very difficult to master and
far removed from any of the living languages. Had this not been the case,
Latin would have probably remained in international usage and eventually
gained a special universal standing. In the period of the industrial revolu-
tion, science, technology, and the arts developed in a number of leading
countries, of which no two had a common language or any similarity in their
languages. This may be the reason why not a single language gained
acceptance as a universal means of communication. In the second half of
the 19th century, some artificial languages were first proposed. The ex-
pectations were very high (Esperanto was considered especially pro-
mising),* but World War I only enhanced the nationalistic barriers and
mankind was too busy to concentrate on the problem of universal communi-
cation. At present, the problem has lost much of its pressing importance
in view of the promising prospects of machine translation, the great ad-
vances in methods of teaching of foreign languages, and the generally
widespread education.

On the other hand, we do not see why a jumble of dialects, languages, and
language families even greater than on Earth need constitute a barrier
against the expansion of a civilization into outer space.

Demographic characteristics of civilization

The statistical data on population growth are fairly accurate only for the
last few centuries. Sufficient data on the proportions of urban and rural
population, distribution according to age groups, etc., in most countries
are available for the last 150 or 200 years only.

The extensive demographic literature either concentrates on the study
and the forecasting of the growth and variation of composition of the
population in individual countries or restricts the treatment to relatively
short periods. Authors analyzing the growth of the Earth population over
the last centuries either do not consider the topic of fitting all the available
data with a single analytical dependence or arbitrarily assume an ex-
ponential function with a correspondingly low accuracy. And yet, the Earth
population data closely follow a hyperbolic function of time, as has been
shown by Shklovskii /10/. Shklovskii has also noted, however, that this

* There are indications that Esperanto is used as a living language in some rural settlements in British
Guiana /9/.

217

EXTRA TERRESTRIAL CIVILIZATIONS

hyperbolic dependence will soon break down as we learn to control
society to a greater extent. The increase of the percentage annual growth
may slow down also because of widespread automation and increased
mechanization, which will make the economic ceiling of a country less
rigidly defined than it was before. Rapid economic growth necessitates
an advanced socioeconomic structure, proper organization and planning
of industry, education, and last, but not least, a sufficiently high rate

of population growth.

The interpretation of various forecasts of the future population of the
Earth may meet with certain difficulties because of the possible discovery
of life forms on the threshold of intelligence /11/ and also because of the
possible creation of near-intelligent systems /2/, teaching of dolphins /12/,
breeding of normal animal species with greater than normal brain capacity
through surgical intervention or selection, or development of man -simulating
machine programs. Forecasting difficulties of another kind are associated
with the future possibilities of suspended animation. Allthese unaccountable
factors naturally introduce a considerable error in the determination
of the Earth population.

In application to other civilizations, the concept of population is even less
certain because of the inherent difficulties in the definition of the concept
"organism." For example, in application to the bees and ants of our world,
some authors treat the entire beehive or anthill as an organism, rather than
a genetic individual /13/.

Social and cultural changes in society may disrupt the quantitative rules
of growth of the entire population and of separate demographic indices.
Thus, the percentage of the human population which inhabited cities with a
population of over 10° varied as follows in different years ¢/14/:

t 1600 1850 1900 1950 1960
p M 2.3 5.5 13.1 20.1

It is readily seen that the growth of this percentage p of the population,
and especially the growth of the ratio of the urban-to-rural population
p /(100—p) does not follow a geometrical progression: the relative growth
rates are distinctly accelerated here.

The data of this table can be approximated with an equation of the form

110
P% = Toreon ^ 100%.

This model can be interpreted as showing a tendency of the entire Earth
(or each continent) to develop into a single large city toward the end of the
20th century. On the other hand, the current tendencies point to a definite
deurbanization, with the urban population migrating to the suburbs. It
would seem that large-scale development of television, a single telephone
and videophone network, remote access to libraries, etc., will retard the

growth of large cities.
The development of individual abilities

The average (or the record) intellectual ability of the individuals and
the extent to which this ability is utilized also provides a certain criterion

5780 218

V. RATES OF DEVELOPMENT OF CIVILIZA TIONS

for evaluating the level and the rates of growth of a civilization. This
analysis, however, involves considerable difficulties, mainly due to lack

of objective and reproducible means for measuring individual ability

and assessing the coverage of these measurements and the choice of signifi-
cant (and independent) characteristics from the entire set of alternatives.

An index which is particularly convenient for measurements may prove
to be ineffective for estimating the growth dynamics if we are unable to
determine its value for the past epochs. So far, no effective methods
have been developed for measuring the general intellectual abilities of men.
The widespread IQ tests yield a numerical scalar index which lumps to-
gether the inherent creative abilities, the sum total of knowledge and
experience, and the cultural level of the person tested. Psychologists are
now ready to admit that these tests are deficient.

In their efforts to pick out the gifted, as well as the knowledgeable,
students, the teachers in some colleges and universities have developed a
certain diagnostic power (intuitive, rather than scientific) which enables
them to identify outstanding creative ability through various educational
competitions.

This intuitive experience, however, is very difficult to apply to estimating
the abilities of the scientists or inventors of the past, when the sum total
of our knowledge, the general picture of the world, and the teaching methods
were entirely different and the people faced problems of a completely
different nature. The present-day population of the Earth is many times
larger than the population in ancient times or in the Middle Ages,
education is much more readily accessible than, say, in Ancient Greece,

Is it not natural to assume that among the modern scientists there are
minds which are at least one order of magnitude more brilliant than Archimedes,
Leibnitz, Newton, and Lomonosov? Is it not likely that Bohr, Wiener, or
Landau would have achieved much more in Descartes’ place? Similar
questions can be raised regarding other fields of human endeavor: poetry
(Pushkin or Blok), prediction of future technology (Roger Bacon or Wells),
chess (Morphy or Alekhin). We can hardly endeavor to answer these
questions without far-reaching advances in the psychology of creative
abilities and in the psychology of constructive education, without going in
minute detail into the particular problems that science and culture faced in
every epoch, and without analyzing the possibilities that each period pre-
sented to people. It is very difficult to find problems which are of the same
difficulty in different periods. The number of foreign languages that can be
learned depends not only on the mind but also on the memory; another
factor to remember is that in different epochs, foreign languages occupied
positions of different importance in elementary education.

There are very few creative problems for which independent solutions
kept cropping up over the ages. The lost proof of Fermat's Great Theorem
could not be reconstructed over three centuries, but how are we to be sure
that the original proof did not contain an error?

Perhaps civilization can be measured in terms of individual
achievements, the extent to which the mental and the physical potential of
the human organism is utilized. Here again, much is still unmeasurable.
For some professions, the productivity of labor has been thoroughly
documented over the ages, but how are we to assess the accomplishments
of a military commander, a teacher, a science writer? Chess masters
can be graded according to the depth and the complexity of the combinations

219

EXTRATERRESTRIAL CIVILIZA TIONS

that they discovered (or missed), but we have no means for estimating to
what extent their ingenuity has been aided by the sum total of the historically
accumulated experience and knowhow.

The various sports are in a more advantageous position in this respect.
But nevertheless, comparison of distant epochs involves difficulties. The
various records are kept only starting with the 19th century; they are
available with high accuracy with full description of the rigidly controlled
conditions. These conditions, however, did not remain constant either:
the footwear and the starting conditions have changed for the sprinter, the
pole-vaulting pole is now made of a different material, etc. The number of new
sports increases rapidly, and there are correspondingly fewer sportsmen
specializing in every given branch (in proportion to the total). On the other
hand, the specialization and the strict professional approach somewhat
increase.

We have briefly considered some aspects of the growth of civilization
which clearly show the difficulties associated with estimating the growth
rates, the variation of growth rates, and establishing quantitative ex-
pressions for the relevant regularities. We are in a much better position,
however, with regard to the rates of development of the other aspects of
civilization, such as economy, technology, and science. These aspects
are of the utmost importance for elucidating the possibilities of space
travel.

In the next two sections, we will consider in greater detail the techno-
logical and the scientific aspects of civilization, but again we will only
present a qualitative description: at this stage, we are more concerned with
the overall picture of the accelerated rate of growth, the conditions under
which the relevant indices describing the development change, and other
topics of this kind. Quantitative characteristics are still available for
relatively small time periods, and it is not at all safe to generalize them to
extraterrestrial civilizations.

$3. INDICES OF TECHNICAL PROGRESS

One of the principal characteristics of a civilization is the level of
technical knowledge, the indices of various technological means, the
quantity and the quality of manufactured products, the amount of energy
used, etc. Another important factor is the proper organization and com-
prehensiveness of the agencies of control governing the entire complex of
technical means and the utilization of technology as a whole.

Science, medicine, and education are also largely dependent on the level
of technological knowledge. At present, however, technology is no
longer merely a means for scientific research, but actually a prime mover
in the advancement of science, constantly challenging scientists with new
problems and tasks.

Extensive literature is currently available on the economic, technical,
and scientific history of individual countries and humanity as a whole.
Numerous studies have been devoted to the growth of individual developmeni
indices (power, velocity, etc. /15,16/). This information, however,
mainly refers to the last decades. The information for the last centuries
is substantially less comprehensive.

220

V. RATES OF DEVELOPMENT OF CIVILIZATIONS

Comparison of numerical data for various epochs of technological
development is not equally valid for all the indices: the growth of energy
output is a more appropriate index than the growth of the pool of metal-
machining lathes, since the productivity and the precision of a lathe
markedly changes with technological development; the capacity of transport
expressed in ton-kilometers is a better index than the number of seagoing
vessels. In general, more valid conclusions can be based on growth indices
whose unit retains a constant value over the various epochs.

The production growth indices, which have been traced over a long period
of time and remain comparable over the entire period, include the world-
wide production of silver or coal, whose rates steadily increased starting in
the 18th or 17th century (when the first more or less accurate quantitative
data were recorded) and up to the end of the 19th century, and then some-
what slowed down. However, coal is now largely replaced by petroleum,
and the rates of growth of petroleum production are much higher than the
present (and past) rates of growth of coal production. Similarly, silver
has been partly replaced by other chemically resistant materials (including
plastics), whose production now increases at rates which were unimaginable
for silver.

In the development of transport, the speed and the power of one form
inevitably reach a certain limit, and then that form of transport is replaced
by a new form which continues developing, and so on.

We thus see that the rates of growth of a civilization are characterized
by a succession of changing leading indices.

On the succession of indices

The above examples show that each stage of development of a civilization
is characterized by certain basic indices which are replaced by other indices
atlater stages. Moreover, we can speak of a certain succession of indices
between the evolution of the animal world and the development of human
society. The development of human society is governed by a characteristic
acceleration of growth rates (and changing of indices) which began in
the course of the biological evolution /17/. It is even possible to set up an
evolutionary scale from 5- 10? years to 50 years, with the successive
periods diminishing by a factor of 10, and compare the principal stages in
the development of life, society, and technology to the divisions of this
scale /18/.

When we pick out a significant index from a number of succeeding
technological means, we are never sure that this index will retain its
significance in the future also.

In the Middle Ages, considerable emphasis was laid on the development
of thermally insulating materials and on the strength of materials in bulk,
whereas later the focus shifted to electrical insulators and the strength per
unit weight (as man progressed from the building of fortresses to the
building of skyscrapers).

Not only the relative role of thé growth of various indices in a given
direction of technological progress changes with time, but the relative
importance of the different directions of progress is also variable. As

221

EX TRATERRESTRIAL CIVILIZATIONS

an example, we can mention the advent of computers and control systems,
which are gradually replacing the power systems as "machines" in the
foreground of technology. On the other hand, the power and energy line
stretches through the entire history of life, and not only technology, on our
planet.

We could continue this extrapolation of ever increasing scales of activity
which engender the succession of the leading (in terms of speed or
significance) aspects of civilization and possibly govern the succession of
periods when the rates of growth are of primary importance for survival
(or in general of relatively high importance) and periods when they are
insignificant (or of relatively low significance). At present, however,
Science does not have sufficiently reliable tools for measuring the level
of the principal aspects of civilization, the rates of growth, and the
significance of their development. Not even the exact function of each
aspect at every stage of development of civilization has been established.
We are not at all certain, for instance, that prehistoric religion ful-
filled any positive function, not even mnemonic, helping to assimilate and
retain in our memory the great variety of empirical factors by dressing
them up in a digestible coat of legend and superstition.

If the growth dynamics of every individual technological means is
nearly exponential, i.e., the percentage annual growth of the corresponding
index is constant, then for an index covering a number of succeeding
technological means, we should take into consideration the change in the
absolute amount of the annual growth. Without special "scaling" of the
time factor in accordance with the general rates of development of techno-
logy and accumulation of technical information, we will simply be unable to
grasp the multitude of numerical data relating to the development of different
branches of technology in different epochs.

Although medicine and education have some features in common with
technology, their growth rate and progress is much more difficult to gauge
and measure. Some of the reasons follow.

The achievements of medicine can be assessed in terms of the mortality
index, the general state of health of the population, and its ablebodiedness.
The state of health and ablebodiedness, however, are not easily expressed
by an objective and reproducible numerical index; moreover, these factors
depend not only on medicine, but also on the socioeconomic conditions,
work and leisure conditions, hereditary predilections of people who have
reached a certain age, etc. It is very difficult to allow for the fact that
the life expectancy of people with some pathological hereditary defects
nowadays is not as different from the life expectancy of normal people as
it was some time in the past.

The difficulties in measuring the progress in the effectiveness and quality
of education are associated with the great diversity of the effects of educa-
tion. Education means knowledge, both applied and abstract, various
"skills," including the skill to apply the acquired knowledge and to acquire
new knowledge. It also means the cultivation of inbred abilities and the
results of training. Social, economic, and technical changes in the life of
society have a prominent effect on the development of children in the school
Stage, as well as on the mind of a specialist after graduation.

222

V. RATES OF DEVELOPMENT OF CIVILIZATIONS

Mathematical functions describing growth rates

We do not give any actual numerical data mainly because the exact
figures relating to the growth of silver production in the world or the
progress in oceanography are of no particular relevance from the point
of view of the search for extraterrestrial civilizations. The rates of
growth of energy and power output have been discussed in Chapters I and
HI. No general conclusions can be drawn regarding the rate of develop-
ment of radio engineering, since the period in question is obviously too
short. 'This reservation is even more applicable to the technology of space
flight. The main purpose of this section is to throw light on the change and
succession of the leading technological indices. Therefore, in practice,
we cannot say in what terms we should characterize the technology of the
supercivilization with which we hope to establish a communication, and
what indices we are to apply to describe its level and rate of growth.

Nevertheless, although we do not intend to give numerical characteristics
of the rates of growth of civilizations, we can offer some comments regard-
ing the mathematical expression of these rates.

A constant growth rate corresponds to a linear dependence of the corre-
sponding index on time. This is very seldom the case for the leading indices.
Most indices characterizing the development of civilizations display rapidly
accelerating growth rates. If the growth rate is proportional to the value
of the index (i.e., the relative growth rate is constant), the corresponding
index increases exponentially with time.

Finally, if the relative growth velocity also increases, i.e., the rate of
change increases sufficiently rapidly with any incremental change in the
index, the index is seen to grow hyperbolically. A characteristic feature
of the hyperbolic law of growth is that the index will rise to infinity in a
finite period of time. Note that numerous indices of growth and develop-
ment of the Earth civilization are adequately fitted with the hyperbolic
curve (e.g., thedemographic index). However, our remarks regarding the
change and succession of the indices indicate that the hyperbolic phase
eventually breaks down for every index.

In a number of cases, the parameters characterizing the development
of civilizations have some intrinsic restrictions (e.g., the ratio of the
number of scientists to the total population). The ratio of growth of this
index naturally increases as it approaches its natural limit. Let n denote
such an index, and suppose that in the early stage of development it
follows the regular exponential dependence

& — const n. (5.1)

At later stages, as the value of the index approaches the limit, the rate of
growth will slow down. We may thus assume that IJ will be proportional

to the difference between the maximum value of the index (e.g., 100%) and
the value at any given time, í.e.,

di
T = const (nmax — n)n. (5.2)

223

EXTRATERRESTRIAL CIVILIZATIONS

Or, in a different form,

wr ea) (5.3)

where fí,is the time scale. The solution of this equation gives the so-called
"logistic" curve

LANA Nmax
no No + (nmax — no) emt? (5. 4)

where mis the initial value of the particular index (for t= 0). The
characteristic feature of the logistic curve is that it is symmetric about
the inflection point. If the curve is not symmetric, it is not a logistic
curve /20/.

Other curves are also used for similar purposes. For example, the
hyperbola can be replaced by Zeman's equation /19/

n — const lg € ; (5.5)
which also leads to accelerated relative rates, but ensures a finite value
of the corresponding integral (i.e., the total number of "events" remains
finite).

In exponential variation, the index always takes a fixed length of time
to double its value. It is for this reason that the growth rates are often
characterized by the time to double the value of the index /20,21/ (see also
next section). For other curves, however, the time of doubling is not an
invariant.

$4. RATES OF GROW TH OF SCIENCE

Science is one of the principal aspects of civilization whose importance
steadily and very rapidly increases. The normal activity of a lathe operator,
or a doctor, gives results which are limited to the immediate neighborhood.
The activity of à scientist, on the other hand, may benefit the whole of
humanity.

On the other hand, ten identical parts turned out by ten lathe operators
are ten times as valuable as a single part. Conversely, ten identical
research projects undertaken by ten scientists are hardly any more valuable
than one of these researches. Science is intrinsically different from
material production, medicine, education, etc., in that, first, each and
every one of its products should be brand new and, second, any product of
Science is not subject to wear and tear under any circumstances, it does
not require maintenance and repair and it can be reproduced at any time in
any quantities. A scientific fact can be used simultaneously in several
places around the world, whereas a lathe is fixed in its location. The pro-
duct of science is not a simple additive sum of all the resources: its effects
are definitely multiplicative in that they themselves act to increase the
resources of all mankind. Another difference between science and
technology is that a scientific product becomes common property, properly

224

V. RATES OF DEVELOPMENT OF CIVILIZATIONS

mechanized and automated, at a substantially later stage of development of
society than an industrial product. The basic requirement of novelty
imposed on every scientific product enhances the importance of chance and
accident in the process of acquisition of scientific knowledge as compared

to the more orderly and systematic nature of material industry and medicine.

Because of these intrinsic differences, science has to be treated
separately from technology, although science is currently gaining in im-
portance as an independent productive force and although scientific research
is often oriented to the immediate needs of industry. The qualitative
characteristics of the growth rate of science therefore merit a section to
themselves.

The main problem in measuring the growth rates of science is the choice
of the significant indices. The number of scientists engaged in research and
the financial allocations do not provide an accurate picture of the role of
Science in society, although these parameters increase very rapidly with
the development of civilization.

Scientific knowledge is the main product of scientific research, and the
growth of this knowledge is a basic criterion of the advancement of science.
However, it is very difficult to form an objective estimate of the amount
of knowledge, and as the main working parameter one generally uses the
number of research workers, the number of scientific publications and
reports.

During the last three hundred years, these indices increase on the
average following an exponential curve /20/, but the time of doubling is
different in different branches of science. In physics, the total number of
publications is doubled in about 10— 15 years /20/, whereas in some
subdivisions of mathematical statistics the corresponding period has
lately been as short as two years /22/. The financial allocations and the
relative number of research workers also increase exponentially, although
eventually this parameter will describe a logistic curve.

The number of creative workers and the proportion of time devoted to
creative and noncreative work by scientists is also variable with time /28/.
Price estimates that the number of really creative scientists is roughly
equal to the square root of the total number of research workers, and it
is they who author approximately half of all the publications and 70— 80%
of the significant results.

At this stage, we can assess the amount of valuable and significant
discoveries in different epochs only by intuition. These intuitive estimates,
although highly subjective, try to guess the number of significant stages in
Science. Feinberg /5/ is of the opinion that the scientific discoveries of
our century are of essentially the same relative significance as the
Scientific discoveries of each of the last three centuries. It is in this way
that he defends the hypothesis of the exponential growth of science. He
points out that only the absolute increment of scientific knowledge, or
Scientific cognizance of the world, increases, and that the growth of this
absolute increment, combined with the fact that in our century science has
finally caught up with and overtaken the role of the arts and humanities in
our society, are responsible for the false impression of the ever increasing
significance of each successive decade, of each successive century, in the
shaping of our scientific knowledge and the scientific picture of the Universe.

225

EXTRATERRESTRIAL CIVILIZA TIONS

However, Feinberg's examples are borrowed mainly from the field of
physics, whose development began earlier than the other sciences; many
of the problems of the microcosmos were solved in the previous centuries.
Physics probably is not the best example for judging the projected rates
of progress of science in general. The significance relations of various
five-decade periods in the history of astronomy and geophysics, chemistry
and biochemistry, physiology and genetics look quite different.

Similar arguments, however, apply to any subjective line of reasoning.
Let us consider the actual growth of the number of scientific discoveries
in some fields of science. The very choice of significant discoveries is
highly subjective, but, once the choice is made, the data can be processed
by fairly reproducible methods. Let us trace the frequency of occurrence
of the dates of various epochs in review monographs dealing with the history
of some branches of science.

The table below gives the frequency of references to various centuries
in one of the fairly popular books on the history of mathematics /25/.
Dates of birth and death of the individual scientists and dates relating to
the development of the history of mathematics (as distinct from the
development of the science of mathematics) were not counted; dates
mentioned in footnotes and in verbal form (not numerical) were also ignored.

Period (centuries) B.C. ist— 14th 15th 16th 17th 18ih 19th 20th
Number of dates

mentioned 30 50 9 19 79 119 255 9
Possible model 8 20 50 125 312.5

The growth from the 15th to the 19th century was close to a geometrical
progression, apart from the exceptional 17th century, the highlight period
in the history of mathematics. The significance of the mathematical
discoveries of the last decades naturally has not received a full evaluation
and we can hardly expect them to appear in a popular book. It is significant,
however, that there are 139 references to the first half of the 19th century,
and only 116 references to the second half. Exponential growth of the
number of dates persists only up to the first half of the 19th century.

Let us now consider the frequency of dates in a source book in the history
of psychology /26/. The dates start back in the fourth century B. C. and the
frequency of references increases steadily up to the early 1930s. Identifying
the years f, in which the cumulative number of references reached an
integral power of 2, i.e., 2Y, we obtain the following table (rounded off to
whole decades):

Y 4 5 6 7 8 9
ty 1580 1670 1780 1850 1890 1930
Doubling time 90 110 70 40 40

The chronological table of the important discoveries and inventions in
the field of chemistry /27/ give the following years as the dates in which
the cumulative number of events reached an integral power of 2 (double

226

V. RATES OF DEVELOPMENT OF CIVILIZATIONS

dates, originallly written with a dash, are replaced by the arithmetic
mean):

Y 2 3 4 5 6 7 8 9 10
50th 22nd 1st 15th 1630 1780 1820 1860 1915
century century century century
B.C. B.C. B.C.
Doubling time 3000 2000 1500 200 150 40 40 55

The last two tables shows that the cumulative number of references has
been growing with a more or less constant doubling time for the last 150 or
200 years only. On the whole, even the progress of the last three-four
centuries cannot be fitted with a single exponential curve. In some books
the cumulative number of references is considerably slowed down starting
as early as the 18th century. This trend is most pronounced in one of the
books on biology /28/. In a number of monographs on the history of
Science, the number of references according to centuries or smaller units
of time increases at a nonuniform rate: for example, the history of
linguistics /29/ shows a sudden upsurge in the number of references in the
16th century, with smaller bursts in the 13th century and the last quarter
of the 19th century.

Thus, our attempts to estimate the volume of knowledge and the
productivity of scientific research at different stages of the development
of science or individual branches of science fail to detect sufficiently
objective and yet significant indices. The number of pages published
in journals is too superficial an indicator, and the number of discoveries
is too subjective.

In each field of science there are probably more reliable and objective
data, such as the measure of completeness, accuracy, and reliability of
the scientific knowledge. A measure of completeness is provided by the
ratio of the number of studied objects of a certain class to the number of
objects of the same class which have not been studied; a measure of
accuracy is provided by the number of significant digits in the results of
measurements of certain parameters, and as a measure of reliability we
may use the length of time needed to detect insufficiencies in the accuracy
of certain parameters. The search for objective indices of the development
of science is still at its embryonic stage.

The subdivision of the aspects of human civilization presented in the last
sections follows the traditional line of reasoning, and is by no means the
best for our purposes; we would be better off operating with more general
terms of stationarity, determinism or regularity in flow and storage of
information, energy, matter, etc. However, a half-baked transition to a
new system of concepts is no more advisable than any attempt to formulate
the final conclusions in traditional terms, which are intrinsically suitable
only for the solution of traditional problems covering much less ground
than the problems discussed in this book. At the present stage of the work
on the problem of progress rates and forecasting in relation to the entire
topic of communication with extraterrestrial civilizations, we are faced
with a necessity of collecting a large volume of various facts, verifying
them, and reclassifying on à new basis. Any attempt to arrive at a

227

EXTRA TERRESTRIAL CIVILIZATIONS

definitive assessment of the interpretation of the facts or even of fact
selection from the standpoint of a particular, restricted conception or a
particular branch of science, whether thermodynamics or semiotics, will
only obstruct future effective approaches to this entirely new and unusual
problem.

$5. FORECASTING

Control systems in nature, industry, and society are equipped for
information acquisition and are capable of classifying the outside stimuli
from the point of view of the required system response, which is intended
to ensure preservation and possibly development of the system itself or of
some larger cybernetic system /30—32/. With some reservations, we can
also discuss output of information from the system. Simple systems evaluate
these outside stimuli only in order to determine the state of the internal and
the external media at the material time, whereas more complex systems can
respond to a forecast future state of the environment as predicted on the
basis of the current measurements. This extrapolatory or forecasting
function of control systems has recently attracted considerable attention in
biology /33/ and in engineering cybernetics /34/.

The importance of forecasts increases as human society reaches
progressively higher levels of complexity and civilization develops. Fore-
casting, in the form of prophecies, was one of the functions of the ancient
tribal chief. Professional oracles and prophets were a common phenomenon
in ancient society, and the truthfulness of their prophecies (e.g., predic-
tion of eclipses) is often attributed to empirical knowledge. The social
recognition enjoyed for a long time by various oracles, astrologists, palm
readers, etc., is associated with their exceptional understanding of human
psychology. Intuitive methods of forecasting, whether truly correct or
simply interpreted as such by the anticipating customer, are generally
erroneously motivated by the configuration of lines on a palm, the position of
planets, or the combination of playing cards, etc. However, the first
advances in modern science completely undermined the prophet's authority
in the educated strata of society. This trend dates back to the theoretical
work of the French materialists in the 18th century. Unfortunately, in
rejecting the parapsychological techniques, they did not investigate the
likelihood of fulfillment of various prophecies relating to the fate of indivi-
duals (it is very difficult to analyze this factor, probably because of the
suggestive influence that such a prophecy may have on the future fate of the
individual), nor did they study the psychological mechanisms of prophecy and
forecasting. During the last 200—300years, the most significant, astonishing
and reliable predictions were made by the leading authorities in each field of
human activity. A correct forecast of the outcome of a military campaign
was expected to come from the military staff or the politicians, a chess
master was regarded as the best authority to offer an opinion on the possible
outcome of a game, During the last decades, the situation slightly changed
because of the closer intercoupling between the various forms of activity and
various fields of human knowledge. The emergence from the stage of narrow
topical specialization of forecasts and predictions inevitably led to some sort

228

V. RATES OF DEVELOPMENT OF CIVILIZATIONS

of a professional specialization and establishment of groups and organiza-
tions whose business is to forecast the future in all fields. On the other
hand, science-fiction writers occupy a progressively more important role
in human thinking. Science-fiction writers have become quite specialized:
Jules Verne wrote a considerable number of pure adventure and travel
novels, without any vestiges of science fiction, whereas today a respectable
Science-fiction writer will hardly write a non-science-fiction novel. Out-
standing scientists acquire a taste for science fiction: H. G. Wells graduated
as a biologist, Isaac Asimov is a well-known biochemist, Arthur Clarke is
an astronomer, I. Efremov is a paleontologist. Some science-fiction writers
eventually switch from literary treatment of their ideas to systematic
analysis of their conception of the future in treatise or monograph form
/35,36/. Subsystems specializing in the forecasting of the future thus again
acquire a special position in the fabric of our civilization.

Classification of forecasts

Forecasts can be divided into those dealing with mass events (which
recur without significant changes in the relevant conditions) and occasional
or unique events (which are very seldom observed, if at all. 'The mass
events are naturally easier to forecast.

There are other possible approaches to the problem of forecasting.

Classification according to the scope of the problem: a) individual or
particular forecasts ("I may lose this peon"), b) forecasts relating to
significant aspects of the fate of an individual, a group of individuals,

a scientific experiment, etc., c) forecasts relating to the development of a
certain branch of industry or science, d) forecasts of the development of thé
entire civilization, e) forecasts purporting to predict the reaction of other
civilizations to the reception of an intelligent signal or the discovery of some
apparatus launched by the originating civilization.

Forecasts are often classified according to the length of their range or
term /21/. Inthis classification, unfortunately, the unit of time is a year
or a century, rather than an epoch defined as the time of doubling of a
significant index. The calendar units used as the exclusive basis for these
forecasts naturally invalidate all comparison of forecasts prepared by a
civilization in periods characterized by different rates of progress or by
different civilizations.

The forecasts can be divided into logically sound and intuitive; another
division is into forecasts using only qualitative data and those based on both
qualitative and quantitative information. This subdivision can be extended
to cover the criteria applied in the selection of the experts for the prepara-
tion of forecasts: the forecasters can be selected on the basis of some
logical tests and criteria or by simple intuition.

In terms of the outcome, forecasts may be deterministic or stochastic;
stochastic forecasts may present a discrete probability distribution, a
continuous distribution of some numerical variable, a distribution in the
function space, etc.

Without going in detail into the systematics of classification of forecasts
and the relationships between the various classifications, we willtry to
consider the means and ways for preparing reliable forecasts and improving
their accuracy.

229

EXTRA TERRESTRIAL CIVILIZATIONS
Accuracy of forecasts

The first step is to learn to compare the accuracy or reliability of
various forecasts or series of forecasts.

The information deficit that the particular forecast failed to
predict or foresee may be used as a quality or reliability criterion for
Stochastic forecasts. Let the possible discrete outcomes K of the forecast
event be assigned the probabilities Px » 0. Suppose that the actual outcome
was Kı. Then -log Px, defines the information deficit: it is zero for Px,=1
and slowly increases to infinity as Px, approaches zero.

Comparing different forecasting methods (or systems) 1, 2,...,j, applied
to events 1,2,..., i,..., n, we draw up the sum of information deficits for

n
each method: S,— — X log Pyix,, S; provides a reliability criterion of the

Series of forecasts obtained by method j: the smaller the probabilities
assigned to the true outcomes, the poorer is the forecast and the higher is
S;. This criterion is additive and convenient in applications. It is readily
generalized to the case of a continuous probability distribution, when the
result of checking the forecast is expressed as some approximate value of
the unknown.

The above criterion can be used for evaluating various modifications of
a forecasting technique, for assessing the qualifications of various experts,
different schools of thought, etc. The criterion can also be applied to
verbally expressed degrees of certainty in different outcomes of a discrete
distribution: the frequency of errors is used to evaluate the probabilities
which are hidden behind such expressions as "possible," "unlikely,"
"impossible," "absolutely impossible."

An expert, having familiarized himself with this error statistics, will
eventually be able to derive stable numerical estimates of probabilities from
such loose expressions, whereas an inexperienced person may easily
interpret these expressions as corresponding to any number between 10%
and 0.01% likelihood of outcome.

If society is interested in evaluating the state of its science and establish-
ing to what extent science is capable of assessing the reliability of hypo-
theses on the basis of indirect evidence, more emphasis should be placed on
polling among the members of the scientific community and on detailed tests
of "intelligent" machines of various types, starting with basic pattern
recognition programs.

It is particularly important to canvass for opinion before the final analysis
of the results of critical experiments, observations of fundamentally new
phenomena, exploratory trips to new parts of the world, of the planetary
system, and other planetary systems. A society keeping a complete record
of the various stages of exploration of the outside world will attain a better
grasp of its own potential.

Since this has never been a common practice on Earth, and the psycho-
logical mechanisms of forecasting have been little studied, we are not ina
position to arrive at a precise and comprehensive comparison of different
forecasting techniques. The helplessness and the ossified approach of
various scientists and science-fiction writers emerges with great clarity
when dealing with far-reaching forecasts of the scientific and technological

230

V. RATES OF DEVELOPMENT OF CIVILIZATIONS

trends of our civilization /36/. Practical failures often intermix with
Scientific inadequacies: both in psychology and in cybernetics, very little
attention has been paid to the basic problem of formulation of a new
hypothesis, as opposed to the selection of the most likely hypotheses
from a given set /37/. And yet, the truly creative activity of mankind
cannot be effectively fitted within the limited framework of the concept of
Selection. This in particular probably provides a partial explanation to the
consistent failure of computers in problem solution, the failure of "heuristic
programming," etc. /38/. To predict in advance the logical likelihood of the
invention of computers or lasers, say, one had to have a very wide grasp of
Sciences and to operate with such abstract concepts as "materials" and
"energy," "control," and "information."

Forecasting the rates of scientific and technological
progress

This particular form of forecasting is better developed than the forecast-
ing of the trends of progress. Examples of far-reaching forecasting of rates
and epochs of scientific and technological progress abound throughout the
history of our civilization.

Some of these forecasts fall wide off the mark, but there are nevertheless
individual valid predictions. The main reason for failure, on the one hand,
is an underestimation of the inherent difficulties of research or invention and,
on the other, an equally dangerous underestimation of the accelerated growth
of science and technology. For example, some ten or fifteen years ago,
scientists did not fully realize the tremendous difficulties of machine trans-
lation or information-theoretical interpretation of the history of languages,
they grossly underestimated the harmful aftereffects of transplantation of
organs or the chemical aftereffects of pesticides. In 1900, H.G. Wells
predicted that atomic energy would be harnessed at the beginning of the
century /39/. On the other hand, he did not foresee the development of the
airplane before the middle of the century. Some seventy or hundred years
ago, some technical achievements, which have by now become a daily reality,
were considered to be impossible (or possible only after millions of years)
/36, 40/.

The simplest method of forecasting the growth of some index is to extra-
polate into the future a theoretical function which closely approximates the:
past empirical growth data. The closeness of the approximation can be
assessed, say, by the least squares method. This approach gives widely
varying results for different horizontal and vertical scales, and the best
Scale is apparently that which gives a function with statistically homogeneous
fluctuations over the entire relevant period of time.

The situation is very uncertain regarding the choice between different
alternative functions when using the least squares method and the determina-
tion of the number of variable parameters for each function.

Some authors stick to the exponential function and the logistic curve /20,
21/, while others use explicitly /10, 16/, or implicitly, to judge from the
Scales of their graphs /17,41/, hyperbolas, exponentials of exponentials,
rational functions, etc. in certain periods of time.

231

EXTRA TERRESTRIAL CIVILIZATIONS

Formal extrapolation of the dynamic series inevitably leads to errors or
to considerable uncertainties in the forecast, if we ignore Such factors as
socioeconomic changes, probable discoveries and inventions, etc. This
severely limits the period of validity of the particular forecasts (development
of individual technical means, narrow branches of science, etc., cf. $2),
whereas the general laws governing the dynamics of rate of growth are
preserved over longer periods /20/.

Psychologically, numerous forecasting errors can be attributed to the
fact that it is not the epoch which is determined by the rate of growth, but
rather the rate of growth is determined by the epoch. A scientist or a
Science-fiction writer estimated the progress in this century using the data
of the last century, and the progress of the current millennium from the
data of the previous millennium, etc. However, the last 300— 400 years are
characterized by a much higher rate of progress than the entire past millen-
nium. There are a number of significant pointers indicating a steady
acceleration of the growth rate over the entire duration of modern history.
The results of this forecasting approach thus lead to an apparent deceleration
of the rate of growth inthefuture, inthatthe future rate of growth is a mirror
reflection of the past growth rates. Consequently, the forecast dates and
epochs will considerably lag behind the actual accelerated development of
Science and technology. Thus, the symmetrical mirror extension of the
sequence of dates 1500, 1800, 1900, 1950 into the future is the sequence 1950,
2000, 2100, 2400, whereas the present acceleration of the growth rates gives
an extrapolated sequence 1950, 1980, 2000, 2015.

There is no evidence to suggest that this last extrapolation is correct, but
the former extrapolation constitutes an extreme case of subjectivism where-
by the current epoch is adopted by the author as the center of symmetry of
the time growth curve.

This subjectivism emerges already on the cover of A, Clarke's book /36/,
where the dates 1800, 1900, 1950, 2000, and 2100 are written one under the
other. The critique of conservative forecasts in this book is restricted to
the psychological level and the level of the actual past history of technology.
Clarke's own forecast of the future growth of technology, although not
accelerated, does not reveal a special accelerated trend. However, the
conception of contracting doubling times is definitely reflected in the fore-
cast growth rates. And yet the currently available future forecasts (see,
e.g., /40/) deal mainly with the prediction of dates and epochs, as does

- Clarke's forecast, and not with the safe growth rates.

The leading importance of the rates of development in any logical forecast
was correctly emphasized by Stine /16/, who unfortunately was carried away
by formal extrapolation, without meaningful analysis. During the six years
after the publication of his work, many of the material growth indices slowed
down (e.g., the increase of transportation speeds). Their hyperbolic growth
gave way to exponential or even slower rates. This correction, ‘however,
does not affect Stine's basic idea, namely that the rates of growth predicted
by Science fiction fall short of the real growth rates: the growth curve of
Science-fiction writers is probably concave from below, whereas Stine
assumes a Straight line.

The forecasting of growth rates of a civilization which has emerged into
outer space, populates a certain part of the galaxy, and continues
developing rapidly encounters specific difficulties associated with the fact

232

V. RATES OF DEVELOPMENT OF CIVILIZA TIONS

that no information can propagate at velocities faster than the velocity of
light, so that there will be a considerable delay in exchange of information
and communication between distant parts of the civilization. In this case,
the communication between the different parts of this galactic civilization
is not unlike the communication between entirely different civilizations.

Forecasting the growth rates of the Earth civilization

Let us now consider some of the topical factors which have bearing on the
forecasts of the future development of humanity. First note that the forecast
growth rates (expressed by some mean curve, rather than a whole family of
curves with probability measures marked for each) are generally determined
for normal conditions: there will be no nuclear war, no lethal microbes will
be imported from other planets, etc. The probability of such a global
holocaust is estimated by some authors to be currently higher than in the
past centuries (when there were instances, if not of total destruction, then
at least of a substantial slow-down in the growth of a civilization, e.g., the
fall of the Roman Empire). However, even if we accept thatthe probability of
a catastrophe has indeed become higher for a particular year or for the life
Span of an individual, it does not mean that this probability is significantly
higher for the entire epoch, since the length of the successive epochs has
shrunk considerably.

On the other hand, the probability of all mankind being wiped out
by some natural catastrophe and the probability of destruction due to external
forces, as opposed to forces operating from within the civilization, has
decreased markedly. We do not foresee a significant danger to mankind as
a result of a sudden fall in the level of solar radiation or the explosion of a
nearby supernova. Since the probability of such events is vanishingly small,
destruction of the civilization due to natural forces is virtually improbable
at the present level of our technology, and external and internal factors
should not be lumped together, as has been done by some authors /42/.

The most common type of forecasts published in the literature is based on
the conception of exponential growth of some index, e.g., power consumption
/10/, or of science and technology as a whole /30/. Is this conception
borne out by the state of things as we face it now, at the end of the 1960's?
There are several significant indications to the contrary.

First, once we assume a certain quantitative dependence of the growth
rate, the index of progress need not remain the same all the time: there may
be a succession of leading characteristic indices describing the development
of the whole civilization or of its individual branches.

Second, as we have seen above, the rate of growth of numerous indices
has been accelerating until recently. "There is no reason to suppose that this
accelerated growth will cease at this particular time. Conversely, it is
more logicalto assume that we are heading for a number of jumps in the
coming years and decades, which will accelerate the rate of development of
our civilization even further. After World War II, the leading countries of
the world invested enormous means in the design of computers, programmed
teaching (and other new teaching techniques), machine translation, elementary
particle research, space exploration. All these investments have so far
yielded only a minor fraction of the expected returns. As a result, there

233

EXTRA TERRESTRIAL CIVILIZA TIONS

are rumors that we are no longer nearing a spectacular jump in these fields.
Are these suspicions wellfounded? After all, space exploration is now
progressing at a very fast rate and has already yielded valuable scientific
results.

New teaching methods have also proved to be highly effective, although on
a limited experimental level only. The problem of machine translation,
however, is still far from its solution, but previous research in this field
has helped to clear the air and to define the various issues connected with
recognition of written messages. Further advances in this field will greatly
promote our understanding of the psychic activity of man and thus lead us
toward a successful solution of numerous problems of teaching and learning,
organization of creative labor and design of "thinking machines."

An important factor in the acceleration of the growth rates of civilization
is obviously associated with the social development of our society.

Decoding of messages from extraterrestrial intelligences will naturally
also enhance the rate of our progress. The rapid growth of radio astronomy
and the projected installation of optical telescopes on the Moon or other
atmosphere-less celestial objects greatly increases the probability of
reception or interception of such messages in the near future. There is also
some hope of detecting traces of technical civilization on the surface of
asteroids and satellites, not subjected to weather erosion.

A third objection against the exponential conception is provided by a
number of circumstances which are liable to slow down the growth of our
civilization in the more distant future (21st or the end of the 20th century).
Some of these factors are purely terrestrial /21/: they are associated with
difficulties of orientation in the growing torrent of scientific information, the
negative effects of the narrow specialization of scientists, etc. Another
fundamental reason is the great difference (by a factor of 10*) in the distances
to the outermost planets of the solar system andtothe neareststars. Because
of this disparity, some authors think that there will be nothing new to conquer
and explore in space for some time after the conquest of the solar system.

A similar situation occurred in the past toward the end of the 19th century,
when the white spots had disappeared from the map and yet no technical
means were available for deep ocean research. This caused a marked slow-
down of the scientific and technical progress, which had accelerated at a very
fast rate before, and contributed to the exceptional popularity of the exponen-
tialmodel. When referring to the rapid acceleration of the growth rates
before the 20th century, I naturally do not mean the annual acceleration but
the acceleration of the periods equal to the doubling time of leading indices
(see $4).

It'is difficult to foresee how the rates of our progress will be affected by
the direct contact with representatives of other civilizations or their auto-
matic machines that is liable to take place in the more distant future.
Science-fiction writers advanced a variety of hypotheses regarding the
impact of this encounter on the scientific and social advancement of the more
backward of the two civilizations. The thesis in /43/ is that no significant
change in the rate of progress can be brought about by this intervention
from outer space unless the recipient society is to lose its individuality.
There is no guarantee that the ideas of enmity and friendship, learning and
exchange of information, observation and experiment are universal, and not
merely anthropomorphic, and that they reflect the entire gamut of complex

5780 234

V. RATES OF DEVELOPMENT OF CIVILIZATIONS

and varied relations between two civilizations. Moreover, once our civiliza-
tion has found its proper place in an infinite system of interrelated and really
friendly civilizations, how are we to be sure that the entire concept of rates
of progress will not prove to be anthropomorphic?

If this is really to happen, then when? No one knows whether this will
take months or millions of years. Humanity started with the idea of the
Earth's unique position in the Universe, and gradually advanced to the
conception of multitutdes of inhabited worlds /44/. At the present stage,
however, we are alltoo acutely aware of the abyssal uncertainty on this
subject. Some authorities believe in the existence of civilized systems in
the Galaxy. Others emphasize that a civilization will hardly need many
millions of years to conquer the entire Galaxy, and anyway the time to
galactic expansion will definitely be much shorter than the entire history of
the Galaxy, so that if an extraterrestrial civilization existed, it is most
likely to have appeared long before the origin of the Earth civilization (after
all, the probability of two twin civilizations is negligible) and would have by
now given signs of its existence. The weak point in this argument is the
implicit assumption of the following factors: the rate of development of any
civilization cannot be consistently (over many millennia) less than the rate of
development of our civilization;* every civilization will be capable of
expanding into outer space; every civilization will expand into outer space.
These implicit assumptions are a reflection of our anthropomorphic chain of
reasoning, and there is definitely no reason to reject the possible existence
of other intelligent beings in our Galaxy or in nearby galaxies.

* The assumption of universal growth rates for all civilizations is vividly expressed in /43/: "We are historians,
not physicists. We measure time in centuries, not seconds... ." In fact, however, history does not retain a
fixed unit of time even for our civilization.

235

Bibliography

1. Macgovan,R. and F.Ordway. Intelligence in the Universe.—
N. Y. Prentice Hall. 1966.

2. Schmeck,H. Semi-Artificial Man. — London, G. G. Harrap and
Co. 1965.

3. Porshnev, B.F.— Vestnik Drevnei Istorii, No.1. A Review. 1963.

4. Sanderson,I. The Dynasty of Abu.—N.Y., Knopf. 1962.

5. Feinberg,E.L. Obyknovennoe i neobychnoe (Ordinary and
Extraordinary).— Novyi Mir, No.8. 1965.

6. Berill,N. Worlds Apart. London. 1965.

7. Kondratov,A.M. Zvuki i znaki (Sounds and Symbols). — Znanie,1966.

8. Frumkina,R. M. Statisticheskie metody izucheniya leksiki
(Statistical Methods of Language Analysis). — "Nauka." 1964.

9, Norwood,V.G.Ch. Man Alone. — London, Boardman. 1958.

10. Shklovskii,I.S. Vselennaya, zhizn', razum (Life and Intelligence
in the Universe). 2nd Edition.— "Nauka." 1965.

11. Porshnev,B.F. Vozmozhna li seichas nauchnaya revolyutsiya v
primatalogii? (Are we Heading for a Scientific Revolution in
Primatology?).— Voprosy Filosofii, No.1. 1966.

12. Bel'kovich,O.M,, S.E.Kleinenberg,and A. V. Yablokov.
Zagadka okeana (The Mystery of the Ocean). — Molodaya Gvardiya.
1965.

13. Chauvin,R. Les sociétés animales de l'abeille au gorille. — Paris.
Plon. 1963.

14. Hauser,P. (Editor). The Study of Urbanization, N. Y. 1965.

15. Rousseau,P. Histoire de la vitesse. — Paris, Press Universitaire
de France. 1946.

16. Stine,G.H. Science Fiction is too Conservative. — Analog Science,
Vol.67. No.3, N. Y. 1961.

17. Meyer,F. L'acceleration evolutive, Paris. 1946.

18. Bruner,J. Toward a Theory of Instruction. Belknap. 1966.

19. Zeman,J. Poznáni a informace. Praha. 1962.

20. Price,O. Little Science, Big Science. — N. Y., Columbia. 1963.

21. Dobrov,G.M. Nauka o nauke (The Science of Science). — Kiev,
"Naukova Dumka," 1966.

22. Nalimov,V.V. and N. A. Chernova. Statisticheskie metody
planirovaniya ekstremal'nykh eksperimentov (Statistical Methods
of Optimum Experiment Planning).— "Nauka." 1965.

23. Lavrentiev,M.A. Berech' vremya uchenogo'. (How to Save the
Scientist's Time). — Organizatsiya i effektinost' nauchnykh
issledovanii, Novosibirsk, "Nauka." 1965.

24. Price,D. Regular Patterns in the Organization of Science. — Organon,
No.2, Warsaw. 1965.

25. Stroik,D.Ya. Kratkii kurs istorii matematiki (A Short Course in the
History of Mathematics). — "Nauka." 1964.

26. Spearman,C. Psychology Down the Ages, Vol.2.— London,
Macmillan. 1937.

27. Walden,P. Chronologische Übersichttabellen. — Berlin, Springer.

EXTRATERRESTRIAL CIVILIZATIONS

1952.

236

28.
29.

30.

31.

32.

33.

34.

35.

36.
37.

38.

39.

40.

41.

42.

43.

44.

V. RATES OF DEVELOPMENT OF CIVILIZA TIONS

Sirks,M.J. and Z. Conway. The Evolution of Biology. N.Y. 1964.

Haus, A. Sprachwissenschaft der Gang ihrer Entwicklung von der
Antike bis zur Gegenwart. Freiburg. 1955.

Lyapunov, A. ÀA.— Conference on Philosophical Aspects of Cybernetics,
Moskva. 1962.

Ashby,W.R. Design for a Brain. 2nd Edition. — New York.

Wiley. 1960.

Bir,St. Kibernetika i upravlenie proizvodstvom (Cybernetics and
Industrial Control). 2nd Edition. — "Nauka." 1965.

Krushinskii,L.V. Ekstrapolyatsionnye refleksy kak elementarnaya
osnova rassudochnoi deyatel'nosti u zhivotnykh (Extrapolation
Reflexes as an Elementary Principle of Decision Making in
Animals). — DAN SSSR, 121(4):762— 765. 1958.

Rubinshtein,S.L. O myshlenii i putyakh ego issledovaniya
(Intelligence and Ways of Its Study). — Izd. AN SSSR. 1958.

Lem,St. Dve evolyutsii — skhodstva i razlichiya (Two Evolutions —
Similar and Dissimilar).— Nauka i Tekhnika, No.8, Riga. 1965.

Clarke,A.C. Profiles of the Future. — Harper and Row. 1962.

Pushkin, V. N. Operativnoe myshlenie v bol'shikh sistemakh
(Functional Intelligence in Large Systems). — "Energiya." 1965.

Pushkin, V.N. Evristika i kibernetika (Heuristics and Cybernetics). =
"Znanie.'' 1965.

Wells,H.G. Anticipations of the Reaction of Mechanical and Scientific
Progress upon Human Life and Thought. — New York and London.
Harper. 1902,

Ryurikov, Yu. Cherez 100 i 1000 let (After 100 and 1000 Years). —
"Iskusstvo." 1961,

Perel'man, R.A. Tseli i puti osvoeniya kosmosa (Means and Aims
of Space Exploration). — "Nauka." 1967.

Zigel,F.Yu. Zhizn' vo Vselennoi (Life in the Universe). — Minsk,
"Nauka i Tekhnika.'' 1966.

Biblioteka sovremennoi fantastiki (Library of Modern Science
Fiction). Vol. 7.—''Molodaya Gvardiya." 1966.

Flammarion,C. Astronomie Populaire.— Paris. C. Marpon et
E. Flammarion. 1881.

237

Chaptev VI

SOME GENERAL TOPICS OF THE PROBLEM
OF EXTRATERRESTRIAL CIVILIZATIONS

$1. INTRODUCTION

A new, scientifically minded approach to the problem of the existence
and development of intelligent beings in the Universe is a definite pos-
Sibility at this stage. There is no need to prove the scientific and the
conceptual importance of further studies in this direction. The question
of the possible existence of extraterrestrial civilizations has cropped up
in one form or another throughout the history of science. This is a
very difficult problem which embraces numerous fields and branches of
Sciences, so that at every particular level of scientific development,
we can tackle only some of the aspects of the problem providing
partial solutions.

The objective materialistic justification for the idea of multiplicity
of inhabited worlds in its original form was the natural desire of man
to penetrate the secrets of the evolution of human beings and human
Society, to reject the theological theses regarding the uniqueness of
human life and intelligence in nature and the intrinsic difference between
"soul" and matter.

Later hypotheses regarding the existence of other intelligent worlds
were advanced in connection with certain scientific and technical advances.
The wider boundaries of the visible Universe, pushed back by the rapid
development of optical astronomy, provided ample food for thought on
the subject of "ecological niches" which could sustain Earth-type life
on other cosmic objects.

From the point of view of modern knowledge, these early hypotheses
probably appear quite naive and unfounded, but the healthy methodological
idea on which these hypotheses were based eventually led to the development
of astrobiology and a new scientific discipline — exobiology, whose task
it is to investigate the possible existence of conditions favoring the
evolution of protein life forms in the Universe.

On the other hand, the problem of extraterrestrial civilizations often
has been considered in connection with forecasts of the future development
of human society. This topic excited the imagination of numerous
Scientists, philosophers, sociologists, and writers. K.E. Tsiolkovskii,
in particular, was the first to point to the possibility of an "energy crisis
with further growth of industry, science, and technology, and he called
attention to the inevitability of exploration and expansion into outer space.

238

VI. GENERAL TOPICS

These ideas, as we know, are adopted as starting premises in the great
majority of modern work dealing with the problem of extraterrestrial
civilizations.

At first glance, the natural and possibly the only approach to the
problem of extraterrestrial civilizations would seem to stem from the
two basic methodological ideas described above, which we shall call the
"exobiological'" and the "predictive" approach. Our reasoning will be
based on the fundamental information available about the protein life
form (the only one observed so far) and on the regular trends in the
development of human society (again, the only known form of civilization
at this stage).

There is, however, a possibility of a more general approach to the
problem of extraterrestrial civilization. In this case, the problem of their
existence is treated as-part of a more general and complex problem which
includes the study of the universal principles of structure, functioning,
and evolution of complex large systems, with biological evolution and human
civilization regarded as particular facets of such systems. This approach
carries fundamentally new methodological features and is directly related
to the development of the general theory of complex systems, which has
recently received a considerable impetus from the direction of theoretical
and technical cybernetics. This approach does not rule out the application
of "exobiological' and "predictive" tools; it actually defines with greater
precision their potential contribution to the problem of extraterrestrial
civilizations.

The idea of a systematic approach to the problem of existence of
extraterrestrial civilizations was clearly formulated by S. Lem /1/,
whose arguments are often quoted in the following.

The problem can be formulated in a slightly different form: should
we not try to analyze even now the fundamental principles of the problem
of extraterrestrial civilizations in order to construct some "general theory
of civilizations" based on the results of modern science? In our opinion,
this is a very real possibility. It will enable us to clearly define the range
of subjects that belong to the problem of extraterrestrial civilizations and
thus define its exact position within the "general theory of civilizations"
and also the position of the "general theory" within the general framework
of classification of modern scientific disciplines.

The significance of the systematic approach clearly emerges from
the examples of the fundamental difficulties which are encountered in the
particular "astronomical" and "radio astronomical" aspects of the problem
of extraterrestrial civilizations.

$2. THE METHODOLOGY OF THE "RADIO ASTRONOMICAL"
ASPECT OF THE PROBLEM. THE "ENERGY" HYPOTHESIS

The advances in radio astronomy place us in a position where we
can reasonably discuss the problems of detection of artificial signals
from space in the radio spectrum. At this stage of our treatment, there
is no need to formulate precise definitions of "signals" and "artificial
origin." The exact meaning of these terms will be clear in each
particular case from what follows, and certain improvements 1. the

239

EXTRATERRESTRIAL CIVILIZATIONS

definition will be introduced in $3 of this chapter. The various ideas on
which the analyses of the radio astronomical'' part of the problem are
based are largely similar to one another. They are reviewed in some
detail in the recent books by Shklovskii /2/, the article by Kardashev /3/,
and in the two comprehensive collections "Extraterrestrial Civilizations"
/4/ and "Interstellar Communication" /5/.

The search for signals from extraterrestrial civilizations is based on
the following fundamental assumptions in these publications:

1) Radio frequencies provide the optimum range for transmission
of meaningful signals over large distances.

2) Civilizations continuously increase their power requirements in the
course of their development.

3) At a certain stage of development, civilizations inevitably start
transmitting information into outer space.

4) The signals received from outer space can be decoded.

Not all of these assumptions are equally valid. "The first of the four
is apparently indisputable. It fully corresponds to the present-day level
of our scientific and technical knowledge. We cannot envisage at this
Stage more efficient and practicable means of communication over
interstellar distances. All other assumptions are highly speculative /2/.

The second point, despite its brief formulation, is characterized
by a complex logical structure. First, it presents a definite prediction
of the future development of human civilization. It is implied that
Scientific and technical progress will steadily continue in the direction
of growing power and energy requirements and conquest of ever larger
regions in space. Second, this principle is extended to all civilizations,
or at least to a wide class of "anthropomorphic" civilizations in the
Universe.*

The predictability of the future trends of scientific and technical
progress is widely discussed in the current scientific press /6,7,8/. The
concensus of opinion is that a comprehensive, systematic approach should
be developed to the various problems of scientific and technical progress,
assisted by a special Scientific apparatus. The analysis should not be
confined to the socioeconomic level of the factors of progress: adequate
attention should be given to the general trends of development, the inner
trends of the evolution of science and technology. This approach gave
rise to a new scientific discipline — the "science of science" (see,

e.g., /8/). One of the aims of this discipline is to devise a general
theory of complex systems.

So far, the best examples of "long-range forecasts" of the development
of human civilization are provided by science fiction writers. An analysis
of the methods and techniques employed in the best products of this genre
yields valuable information on the "psychology of forecasting." The main
feature of these methods is the application of linear extrapolation into
the future of those factors which are currently being implemented or are
potentially ripe for implementation in the near future. S. Lem calls this
technique "orthoevolutionary forecast." Direct time tests show that even

* The last sentence requires some qualification. Jt will be seen from the following that this statement /3
may be interpicted to heve a two-fold meaning. On the onc hand. it may imply that all thc extra-
terrestrial civilizarions are anthropomorphic, and on the other hand, «c may intentionally limit the
discussion to anthropomorphic civilizations.

240

VI, GENERAL TOPICS

the forecasts for the relatively near future greatly differ from the actual
reality. The dialectics of growth, as we know, includes the quantitative
changes as one (by no means principal) stage in the evolutionary process.
In certain periods of development of human society, the discovery

of new horizons (e. g., new forms of energy, new materials) produced a
leap-like qualitative change in the methods of production, largely altering
the way of life of the current generation and the further development

of the society as a whole. Another shortcoming of the "orthoevolutionary
method" is that it does not predict any alternative courses of development.
One aspect of the phenomenon being considered is treated as an absolute
factor, which is placed in antagonistic opposition to all the other alter-
natives. Actual evolution, on theother hand, is molded by an incessant
interaction of polarities, and this interaction is one of the basic principles
of the dynamics of progress.

A weak side of the "energy" hypothesis of the evolution of human
civilization is its pronounced "orthoevolutionary" character. "It is tacitly
implied that the rate of growth of the technical progress observed on the
Earth during the last 200 years is a dynamically continuous process which
can be arrested only by violent destructive forces ("degeneracy" or
"suicide" of a civilization)" (/1/, p.85).

The basic premises of the "energy" hypothesis are clearly based on
known facts in the history of the development of science and technology
during the recent period. The main motivation of this idea is the healthy
desire to foresee and avoid the dangers of the forthcoming "energy" or
"demographic" crisis, predicted in various sources /6,7/. The conception
of continuous expansion throughout outer space follows directly from
the basic premises of the "energy" hypothesis and does not constitute a new
additionalassumption. However, the only cure that the "energy" hypothesis
prescribes for these crises is a further quantitative step-up of power output.
The search for new power sources and "lebensraum" is thus elevated to
the pedestal of eternal problems. It is postulated that the current charac-
teristics of the dynamic growth of humanity will persist for an indefinitely
long period of time (in fact, hypotheses of this class maintain that this was
the course of civilization from its very inception).* 'This principle
applied to the energy hypothesis leads to the conclusion that the search for
new power sources and free space places humanity in the uncomfortable
position of striving to balance itself on a "razor's edge," since the slightest
delay in making new power resources available will lead to catastrophic
results. This explains the great importance attached to "space
engineering" projects (Dyson's sphere, for instance) which constitute models
capable of resolving the power and demographic crises. However, unlimited

" Note that, following in the steps of the originator of the idea of man's expansion into outer -pace,
K. E. Tsiolkovskii, the Soviet authors generally associate this trend with their optimistic ethical-
philosophical confidence in the unlimited potential of the human mind. On the other hand, Western
scientists reveal a tendency to interpret the exodus into outer :pace as a result of the hardships of
life and the conflicts of modern society. R. Simon /9/, for instance, tries to promote outer space,
in the best Madison Avenue style, as a marvellous place for the development of private enterprise
which is severely hampered on Earth. The main advantage of outer space, according to Simon, is
its enormous "rhrce-dimcnsional capacity." Therefore the "excess numbers" of humanity will spread
to other planets, in order to escape from the congestion of the Earth, and thus enhance the "harmony"
of human society.

241

EXTRATERRESTRIAL CIVILIZATIONS

"diffusion through space" and increase of power output increases the pro-
bability of other crises, which can be envisaged already at the present stage.
We mean here the "information" and "organization" crises, which are
ignored within the framework of the "energy" hypothesis. The authors

who criticized the "space engineering" approach specifically mentioned
obstacles of this kind. On the other hand, a detailed analysis of the means
for overcoming these effects in the course of evolution leads to hypotheses
which are radically different from the "energy" hypothesis /1/.

To summarize the preceding arguments, we would like to stress that
the "energy" hypothesis is one of several alternatives based on the analysis
of certain tendencies in the current development of humanity.

Let us now consider the universal applicability of the "energy" approach
to the growth of other extraterrestrial civilizations. The Earth civilization,
the only actual example before us, grows "technologically."* In theory,
however, we should not reject the possibility of a 'nontechnological"
growth of a complex animate system /1/. A typical example of such
growth is biological evolution, which takes the course of plastic
adaptation to the environment. A system which develops in this way may
reach a very high level of organization. From our point of view, however,
it is not "intelligent." This conclusion is a fact for the biological evolution
observed in this world. We can visualize, however, a directional activity
taking the form of programmed autoevolution and introducing biological
modifications intended to improve the adaptability to the environment.

This civilization would appear very odd indeed from the point of view of
the Earth civilization. This oddity, however, may beadirect result of our
anthropocentric way of thinking, which automatically rejects the possible
existence of other intelligent forms.

In any case, the definition of an "intelligent" system and to what extent
it may be regarded as a "civilization" requires additional analysis.

The "energy" hypothesis prescribes one universal course of development
for all the "technological," "anthopomorphic"' civilizations.

Let us now consider the third basic assumption contained in the
hypothesis regarding the feasibility of "radio astronomical" detection of
signals from an extraterrestrial civilization. On the one hand, the third
assumption is necessary to ensure a logical closure of the problem. If
civilizations do not transmit radio signals, they cannot be detected by
means of radio observations.** On the other hand, the third assumption
nevertheless requires some logical justification.

We will naturally consider the case of a civilization intentionally
transmitting information into outer space. This activity of extra-
terrestrial civilizations is generally justified on two counts:

a) it is assumed that the transmission of signals is related to the
experimental, exploratory activity of an advanced "technological"
civilization, trying to locate other similar civilizations /2/;

* Humanity took the course of active modification of nature. creating suitable conditions to sustain the
parameters which are needed for its existence. This activity is far from "harmonic." Willingly or
unwillingly, man destroys the natural ecological balance hy thi» intervention. creating an artificial
“ultra-low entropy” environment.

** This conclusion is fully valid in the "anthropocentric" statement of the problem. In general. we should
consider ways and mcans for detection of civilizations which do not transmit special signals announcing
their existence but which arc nevertheless "manifested" in specific forms of "behavior" (see £3).

242

VI. GENERAL TOPICS

b) it is assumed that the psychological-ethical trends of a highly
organized civilization generate a certain pressure for the transmission of
signals into outer space /3/.

Arguments of the first group clearly stem from the "energy" hypothesis
applied to the growth of a "technological" civilization. "Technological"
civilizations must explore the entire gamut of natural effects in the entire
Universe around them. The most logical instrument for this exploration
would be to establish a communication channel between neighboring
civilizations. The "delay" in bilateral communication due to the tre-
mendous distances in outer space does not constitute a fundamental
difficulty in this treatment.

The second group of arguments is also sometimes quoted to justify
the "voluntary" transmission of information to an unknown receiver.

This factor is highly significant in calculations of the probability of
detection of signals from supercivilizations /3/.

It can be argued that any functional activity of any complex system is
justified only if it is essential for healthy growth and development of the
System. In this sense, ethical and psychological factors are an outgrowth
of deeper "behavioral principles" of a civilization. Therefore, the usual
approach to the humanism or, conversely, the "aggression" of a civilization
constitutes an entirely new factor added to the long line of previous
assumptions, based on the extrapolation of current notions and concepts.

Let us now consider the possibility of decoding of the received signal.
The logic behind the earlier attempts leads to the conclusion that an
"anthropomorphic," "technological" civilization should transmit information
in the form a semantic language system encoded in a certain form.
Numerous attempts to construct formal languages for transmission of
anthropomorphic concepts are known. One of these is Freudental's LINCOS
/10/. The decoding approach is based on the assumption that the general
system of concepts and knowledge is the same for the communicating
civilizations. This assumption, however, is not a logical outcome of the
preceding treatment. Gladkii /11/ pointed to the theoretical possibility
of the existence of systems of knowledge with radically different elementary
concepts for different civilizations. Inthese cases, the decoding of messages
will naturally encounter serious difficulties.

For the purposes of our treatment, we should emphasize that the "radio
astronomical' hypotheses of search for extraterrestrial civilizations are
again committed to the anthropomorphic approach in their signal decoding
attempts.

The analysis of assumptions which constitute the basis of /3/ and a
similar group of hypotheses shows that the object of our search are signals
from extraterrestrial civilizations which are close to the Earth civilization
in all their basic activities and manifestations (including the details of the
forecast growth).

It is naturally very difficult to follow blindly the "pure" principles of the
form advanced in /3/. After all, we are dealing with working hypotheses
and even their authors themselves continuousiy search for new ideas and
methods of search for signals from extraterrestrial civilizations. In this
sense, the aim of our critique is not to "reject" the current assumptions,
but rather to define clearly the methodology behind these hypotheses.

243

EXTRATERRESTRIAL CIVILIZATIONS

A transitional stage toward new constructive possibilities of investigation
is provided by the comprehensive discussion of the artificiality criteria of
signals from outer space. Shklovskii proposed the conception of a "cosmic
wonder" /2/. By "cosmic wonder" he understands the manifestations of
intelligent activity on a cosmic scale, as observed by astronomical methods.
Within the framework of the radio astronomical search for signals from
extraterrestrial civilizations, the problem of discovery of the "cosmic
wonder" sounds similar to the discovery of "call signals," i. e., signals
which carry explicit information pointing to their artificial origin (see
Chapter III). There is, of course, the question of unambiguous inter-
pretation of this effect. For example, the decoding of a certain semantic
system of signals received in the radio spectrum from an astronomical
object would clearly point to the existence of a transmitting civilization.
The artificiality criterion in this case was provided by the very decoding
of the information contained in the signal. This, however, is an extremely
lucky and quite unlikely turn of events. What are the other alternatives?
According to Shklovskii, "... wecanoftendetect distant supercivilizations
purely objectively, by observations, because the associated objects do not
follow the laws governing the behavior of inanimate matter or display
remarkable and probably unnatural characteristics" /2/.

Statements of the kind ''do not follow the laws governing the behavior
of inanimate matter' and "remarkable characteristics" are ambiguous
and uncertain. The logic behind the development of the natural sciences
and their application to the study of astronomical objects imposes certain
rigid restrictions on the possible interpretation of the most "outlandish"
phenomena in the Universe as manifestations of "intelligent" activity.

Lack of precise criteria which distinguish the product of activity of a
civilization from natural cosmic objects is conducive to unscientific
Speculations regarding the artificial origin of certain unusual phenomena
(a well-known example is the unfortunate notoriety of the Tunguska meteorite)
and, on the other hand, imposes an unnecessary "restriction" on the
discovery and astronomical investigation of fundamentally new "natural"
effects which enrich our knowledge in the fields of physics and other
Sciences.

A systematic approach to the solution of the problem of artificiality
criteria will be described in $3. Meanwhile we will consider another
topic which is still the subject of lively discussion in the scientific
literature concerned with extraterrestrial civilizations.

An important experimental fact is the conspicuous absence of
"cosmic wonders." At this stage, we will ignore the possibility that this
is due to our unreliable "artificiality" criteria and adopt the "anthropomor-
phic" approach. In accordance with the "energy" hypothesis, all civili-
zations should pass through a "technological" phase in their development,
exploring and conquering the surrounding space. Many authors have noted
(see Chapter I) that the rate of this technological development should be
very high. In practice this means that the manifestations of cosmic activity
of civilizations will be noticeable over periods which are very brief
compared to the cosmic time scale. The absence of cosmic wonders
within the framework of our hypotheses can therefore be attributed a)
to the extreme rarity of civilizations, b) to the fact that all civilizations
are roughly in the same "early" stage of development, and c) to the
relatively short lifetime of civilizations. S. Lem examined these three
possibilities critically.

244

VI. GENERAL TOPICS

What is the evidence in favor of the extreme rarity of civilizations?
Baumshtein /12/ tried to prove the uniqueness of life on Earth by applying
probabilistic and combinatorial techniques to calculate the likelihood
of various ancillary conditions necessary for bio- and anthropogenesis
(a certain gravitational pull of the Moon, a "required" succession of
climatic conditions, etc.). S. Lem /1/ justly criticized the validity
of the application of combinatorial methods to the highly complex dynamic
evolutionary system. The evolutionary system is fundamentally plastic,
so that the presence or absence of certain secondary conditions does not
present it with the binary choice between "life and death" and only imposes
certain restrictive trends on its future development. These calculations
only prove the extreme unlikelihood of the development of an exact replica
of the terrestrial anthropogenesis on other planets and thus place all the
"anthropomorphic' hypotheses on shaky ground.

The words "extreme rarity" (if the other alternatives are ruled out)
indicate an extreme dispersion of civilizations in the Universe, so that the
Earth civilization is the only one occupying almost the entire visible
Universe (otherwise, we would have witnessed the activity of super-
civilizations of Kardashev's type). This assumption should be made
consistent with the accepted cosmogonic concepts regarding the repre-
sentativeness of the conditions in the solar system, the representativeness
of the solar system in the Galaxy, and the representativeness of our
Galaxy in the Metagalaxy.

The "rarity" of civilizations, however, is not sufficient to explain
the total lack of observational evidence of their activity, unless we assume
that the Earth civilization is unique.

Assumption (b) seems to contradict the cosmological information
regarding the different age of the cosmic objects and the evolution of the
Galaxy and the Metagalaxy. We can hardly accept the suggestion that the
conditions favoring the evolution of life arose relatively recently on tne
cosmic time scale throughout the observable part of the Universe.

To justify the assumption of the brief lifetime of civilizations, various
authors generally speculate about the possibility of catastrophes and crises
which emerge in the course of accelerated growth /13/. However, the logic
of these ideas in application to the conspicuous absence of "cosmic wonders"
leads to the inevitable conclusion that all civilizations unavoidably perish
in the early stages of their "technological" development /1/. It is only by
adopting this fatalistic point of view that we can understand the lack of any
signs of activity of "surviving" and rapidly advancing civilizations. This
monstrous determinism is very difficult to accept without questioning. *

* We should emphasize that the idea of all civilizations perishing at a "convenient" moment, i.e., on the
threshold of emerging into outer space or right before this phase, is particularly unlikely. In principle,
the “death” of an individual civilization is far from contradicting the basic premises of the dialectic
philosophy, which postulates that only matter as a whole is “indestructible,” while all other phcnomena
originate. grow, and perish, giving room to new forms of life and existence /12/. A contrary point of
view would have led to the erroneous statements that “the Universe is permeated with intelligence" or
“intelligence is an indestructible attribute of matter.” If we were to adopt these ideas, the lack of
"cosmic wonders” would again lead us to the conclusion that the Earth civilization is unique in the entire
observable Universe.

It is therefore tnore logical to assume that highly organized forms of existence regularly develop in
different comers of the Universe. The spark of new life burns brightly, only to become extinguished and
then reborn again under appropriate conditions. The span of life of these systems is difficult to predict.
Modem science, in our opinion, does not provide even a rough estimate of the duration of a typical
"psychozoic" era.

245

EXTRATERRESTRIAL CIVILIZATIONS

Proceeding from the absence of apparent signs of activity of extra-
terrestrial civilizations and the inherent weakness of the "anthropomorphic"
hypotheses, S. Lem advanced an interesting hypothesis which maintains
that the "nontechnological" evolution is characteristic for most existing
extraterrestrial civilizations. According to S. Lem, the current "energetic"
phase, including the expansion into outer space, constitutes only a very
brief period in the life of a civilization, and it will eventually be replaced
(in particular, under pressure from "information," "organization," and
other crises) by a qualitatively new form of evolution. Lem's hypothesis
naturally accounts for the absence of "cosmic wonders" and does not seek
a definite answer to the question of the lifetime of civilizations. Note that
from the traditional point of view we can hardly accept the idea ofa
"restriction" imposed on the expansion of a developing civilization into
outer space. In any case, Lem's hypothesis contains fewer internal
inconsistencies than any "anthropomorphic" hypothesis.

Let us now try to generalize our analysis of the weak sides of the
theories of existence of extraterrestrial civilizations and the problems
of communication based on "exobiological" and "predictive" principles.

A significant methodological shortcoming of the hypotheses of this
group stems from an excessive abundance of additional assumptions.

These a priori assumptions are associated with the "orthoevolutionary"
reasoning of the authors (forecasting of future development based on

linear extrapolation and "anthropomorphism,' which maintains that this
mode of growth is applicable to all (or most) extraterrestrial civilizations.
According to this approach, certain facets of the phenomenon are invested
with absolute importance and decisive significance, whereas other possibilities
are ignored (e. g., the consequences of an "information" crisis).

As a result, theories developed in this way cannot resolve eventhe
difficulties associated with those effects which are taken into consideration.
Thus, for instance, the problem of the "energy" crisis is removed to
outer space and its solution is postponed until "better times."

The "anthropomorphism" of our conceptions prevented a satisfactory
development of the highly important notion of the "cosmic wonder." The
artificiality criteria can be derived only from an internal system of
anthropomorphic concepts. Therefore, the only clearcut criterion within
this framework is a literal, word-by-word replication of external mani-
festations of human activity on other cosmic objects (e. g., adoption of
an "anthropomorphic" semantic system of signals).

At the same time, the basis for the extrapolation of anthropomorphic
hypotheses is provided by our knowledge of the protein form of life and the
structure of the Earth civilization at the present phase of its development.
Is there not an alternative course leading to a less controversial and more
conclusive "theory" of extraterrestrial civilizations?

Any nonspeculative hypothesis should clearly rest on a foundation of
scientific data. In this sense, we can maintain that the solution to the
problem of extraterrestrial civilizations should be sought (at least at this
stage) on Earth! Will it be enough to allow for all the present-day scientific
concepts in the construction of this theory? Shklovskii /2/ pointed to the
fundamental importance in the theory of extraterrestrial civilizations of such
half-baked topics as functional definition of life and "intelligence."
Shklovskii's argument sounds as if the respective studies are still in

246

Vi. GENERAL TOPICS

their embryonic stage and have not yielded any tangible results which can
be applíed to the problem of existence and forms of intelligence in the
Universe.

The theory of extraterrestrial civilizations is naturally in great need
of exact definitions of life and intelligence, which are not restricted by
any narrow particular model and the distinctive features of bio- and
anthropogenesis. It is moreover clear that we are still very far from the
development of sufficiently clear definitions of this kind. On the other hand,
modern scientific disciplines related to cybernetic techniques indicate a
new approach to the investigation of the surrounding reality. The funda-
mental nature of this novelty opens wide horizons in front of the correspond-
ing branches of human knowledge. The new method of "cybernetic intel-
ligence" is highly effective in problems dealing with the study of complex
large systems.

The “cybernetic methods" enable us to introduce some order in the
problem of extraterrestrial civilizations, to refine the terminology, to
estimate the objectivity of the various statements, and finally to come
up with a correct formulation of the basic problems. In what follows,
we will try to present a systematic description of some constructive
principles guiding the application of this method to the problem of extra-
terrestrial civilizations. We will also try to revise our attitude toward
the various assumptions of the earlier theories. First, however, we
will consider those publications in which the systematic approach has in
fact been applied.

$3. AN ALTERNATIVE POINT OF VIEW. S. LEM
AND HIS SUMMA TECHNOLOGIAE

Inhis book, Summa technologiae/1/, S.Lemtreatedindetaila number
of topics associated with the problem of extraterrestrial civilizations. The
"astronomical" aspect of the problem received only minor attention in this
book. We have tried to show, however, that the problem of the existence
of extraterrestrial civilizations is in fact part of a much wider problem
concerned with the properties and the evolution of highly complex systems.

S. Lem deals with these "adjoining" problems and concentrates mainly
on the possibilities of "forecasting" the future growth of civilizations. The
principal features of the biological evolution are examined in detail in the
light of modern scientific data. The potential possibilities of natural
biogenesis and the control of "live" systems are discussed. These pos-
Sibilities are compared with the requirements presented by science in
connection with the design of complex artificial systems.

In /1/ it is noted that all the theoretical constructions dealing with the
"forecasting' of the future development of mankind somewhat idealize the
"thoroughness" of biogenesis. This is so because at the present stage
mankind is still incapable of reaching the same degree of perfection in
organic synthesis as the natural biological evolution has reached. We
therefore tend to attach absolute importance to the gifts of nature, ignoring
any possibilities of an "improvement" of human nature by artificial
"autoevolution." "When chemical synthesis, the theory of information, and
the general theory of systems reach a highly advanced stage, the human

247

EXTRA TERRESTRIAL CIVILIZATIONS

body will appear in the light of these achievements as the most imperfect
element." The next step naturally will be to seek ways and means for
improving the "least perfect" element in the system of civilization. This
will probably be achieved by "autoevolution," and not by improvement of the
"conditions of life," since the very imperfection of the natural organization
of the human organism sets a fixed limit to human life at around 100 years.
It is frequently suggested that the human life span can be easily stretched
to 150 — 200 years by appropriate medical treatment, but we doubt that this
is indeed so. The basis for this suggestion stems primarily from the
belief in the "enormous potential possibilities" presumably hidden in the
human organism. In the process of evolution, the organisms reach a high
degree of plasticity, adaptability, and build up a "reserve of reliability."
The biological evolution, however, does not "plan" or "provide' for the
future. Only the fittest survive, i.e., those organisms which are best
adapted to the existing conditions.

Only those features of the biological evolution are selected which are
significant for the ultimate "purpose" of genesis, i.e., for the survival
of the species as a whole. As regards longevity, nature is "not interested"
in the fate of the individual or how long he lives after having fulfilled the
life functions significant for the continuation of the system (procreation,
guardianship of the young generation). Accidental factors may combine
to enable one individual to exceed a certain "necessary limit" of life
expectancy. "Anthropomorphism" in the approach to the biological longevity
of the human organism presoribes imaginary potential reserves to the bio-
evolution, which presumably lie hidden until they are needed. In the final
analysis this leads to teleological views on biogenesis. Moreover, the
"methods" of evolution are purely statistical. The fact that certain in-
dividuals live beyond an average maximum age does not prove the longevity
of the species as a whole. The fact that some people live to the age of
150— 170 years is analogous to the fact that some people are geniuses:
any suggestion that the entire population can be "educated to the level of
genius" by an improvement ofthe conditions of life cannot be taken seriously.

An important place in S. Lem's book is devoted to the analysis of the
"information" crisis. The danger of this crisis is considered in conjunction
with the current trends in the development of science. The accelerated
growth of the production resources (in particular, the search for new power
sources) requires an ever increasing quantity of scientific information.
This trend emerges from an historical example, analyzing the amount of
research which had to be completed to ensure a transition from one kind
of power to another. The corresponding amount of research steadily
increased. The development of modern society will stop if the rate of
data acquisition will cease accelerating. On the other hand, the "necessary"
Scientific discoveries cannot be planned or programmed. The "strategy"
of science is essentially a matter of chance. In the course of scientific
progress, the Earth civilization allocates scientific efforts to all the possible
fields of research, since we do not know beforehand what turn the funda-
mental discoveries will take. This state of things naturally leads to an
avalanche of new information, and also ties up a progressively larger
number of people in scientific research. The high rate of growth should
inevitably lead (in the light of quantitative treatment of the problem) to
catastrophic results (e. g., depletion of human reserves that can be tapped
by the needs of scientific research) The random character of the

248

VI, GENERAL TOPICS

"generation" of important discoveries prevents us from imposing any
reasonable restriction on the scope of research. Given the present state
of things, this would lead to even more serious consequences. Lem notes
that even today excessive hyperbolization of individual fields of research,
associated with rocket engineering and space exploration (the socio-
political reasons are not to be ignored here), has a detrimental effect on
basic research in other fields. And yet, different branches of science
are not isolated or independent. Therefore, artificial retardation of the
growth of some branches will eventually produce serious interference with
further progress in the privileged branches of science also. Hypertrophy
of the "popular" sciences often leads the scientist to lose much of the
finesse of research and to try to solve all knotty problems by "frontal
attack," by sheer quantitative step-up ofthe power level of the experiment
(ever more powerful particle accelerators, ever larger radio telescopes, etc. )
/14/.

How are we to avoid the "information" crisis? The development of
science in human society reveals the particular importance of the ever
increasing "information channel" between nature and the civilization.

So far, the "channel" has been broadened by adding new branches of
science and ever larger numbers of research workers. The advances

in cybernetics give grounds for hope that in the near future we will be able
to design complex machines to help man in the acquisition and processing
of the increasing quantity of information. This does not indicate a change
in the general character of scientific research, but only "automation"

of the process.

It is clear, however, that by relying on "synthetic scientists" we only
postpone the imminent crisis, without actually liquidating the factors
responsible for the entire development. The only option (all other condi-
tions being constant), according to S. Lem, is to create an artificial
system which would directly extract the relevant information from the
environment, i.e., a machine which would act not as a mere assistant for
information processing, but as a powerful analytical system with capa-
bilities far beyond those of the human brain. Lem develops the con-
ception of a certain synthetic evolutionary system capable of increasing
the "output" information in the process of its development, i.e., an
arrangement not unlike the accumulation of genetic information in bio-
genesis, but with "directional and improved" action. A system of this
kind will "cultivate" scientific results and conclusions. Lem shows that
the feasibility of such a system does not contradict the premises of modern
science and he proceeds to analyze the design details of the system from
the point of view of the mechanism of scientific cognizance and the means
of information transmission in biogenesis.

However, the creation of an autonomous data processing system will
solve only part of the problems. The "information" crisis is not an
independent phenomenon: it is conditioned on a whole range of other
important satellite processes.

Lem also discusses the likelihood of other crises and catastrophic
developments. Even if the problem of information acquisition finds a
satisfactory solution, we will probably be far from a "quiet' mode of
development in the form of "colonization of outer space" with gradual

249

EXTRATERRESTRIAL CIVILIZATIONS

expansion of the "living" space, while the population and the power resour-
ces will keep increasing continuously. One of the basic principles of
existence and activity of complex systems is their controllability. We
are not aware at this stage of the existence of a definite limit of structural
complexity of a system (number of component elements). There isa
possibility, however, that when the number of component elements

exceeds a certain critical value, the system will become uncontrollable
and disintegrate. In application to the "overgrown' civilization of the
future, the problem of control no longer reduces to the banal and half-
jocular question "how do we keep all the members of society busy?" We
are dealing with such fundamental aspects, as, say, preserving the culiural
unity of the giant system. An important cementing link in the development
of humanity is the very continuity of the various stages, the transmission
of "information" from generation to generation through vigorous exchange
between the individual members of society. This process naturally en-
larges the horizons of every individual. The impediment of information
exchange in a giant supercivilization may lead to a loss of individuality,
and every member will face the danger of becoming a highly specialized
"cell" fulfilling narrow and limited service functions (no cther "rational
controllability" of the giant system can be visualized).

Lem notes that the concept of civilization is far from being synonymous
with the free growth of all the possible individual freedoms. The opposite
is probably true: the development of society imposes ever new restrictions,
which are a necessary evil. This leads us to an interesting question:
supposing scientific analysis confirms the unfeasibility of controllable
systems made up of an excessive number of elements or shows that such
a giant system can be rendered controllable only by bringing all the
members of society to one common level, will this not be an excessive
price to pay for the "freedom" of unlimited expansion into outer space?

Theoretically, we can reasonably assume that a society which has
encountered fundamental difficulties on the way to expansion into outer
space will reject this course of development. This need not indicate
any "degradation" of the civilization. It has never been proved that the
"spontaneous" growth of human civilization is the "best natural course"
and will never lead to negative results or to tremendous irrational expen-
ditures in the future. It suffices to mention the undesirable effects of the
recent uses of atomic energy, such as the danger of genetic degeneration,
or the harmful consequences of the wild, "unhusbanded'' dissipation of the
natural resources of the Earth.

Once recognized, the need is readily accepted by the civilization and
is never regarded as an "unnatural" or "contra-natural' factor. For
example, the "demographic" crisis is on the whole easily solved within the
framework of ethics by birth control and family planning (a technique which
is becoming progressively more popular with the growth of materialistic
culture).

As an alternative to the "orthoevolutionary," "energetic" forecast of
development, Lem advances a different hypothesis of his own. We have
already mentioned the concept of "autoevolution." Another possibility,
which enhances the "autoevolutionary" trend in Lem's opinion, is the
creation of a "world within a world," i.e., a conglomerate of artificial
conditions in a sufficiently large volume of space which is governed by a
System of programmed artificial conditions characteristic of that "world."

250

VI. GENERAL TOPICS

The laws of motion, signal propagation, and structural elements of all the
material objects in this "reserve" should be chosen so that they ensure
optimum "control" of all the objects in that world, of the "imprisoned"
civilization. This system would comprise an artificial machine consisting
of two basic parts: the "environment" and the "civilization." The
"function" of the machine amounts to the interaction between the two
component parts, Lem in his Summa technologiae analyzes the
feasibility of such systems.

Lem's hypothesis is of course highly speculative. However, it may
lead to important methodological conclusions: besides the "energy"
approach, there are alternative courses that a civilization may take,
which theoretically are no less probable. The underlying idea of this
treatment is that if the operating principles of complex systems can be
disclosed, sooner or later the scientific progress will enable us to
identify the optimum modes of development. Lem thus reduces the
problem to its elemental level: what should be the "aim" of a civilization
and what course of development should the civilization take in order
to achieve that "aim"?

The "aim" of a complex system can be interpreted as the internally
recognized principle of its action. We can speak of the objective categories
of "intelligence," "conscience," "logic of systems," emphasizing that they
are all functional properties of complex systems. The emergence of a
certain "metaphysics' or "dogmatism" is inseparably linked with the practical
activity of a complex system identified as a civilization. At every stage,
the system does not have "complete knowledge" of the environmental
reality, but it nevertheless functions as if its knowledge were complete.
The functional determinism is associated with the conviction of its
"correctness," as otherwise the system simply would not function. This
is the basis of the natural "metaphysics' or ''dogmatism." In the course
of its development, a civilization progresses through a long succession
of "dogmatic" or "working" hypotheses, which govern its functions and
constitute a certain approximation to the objective reality. The process
degenerates into "pure dogma" if the constant experimental checking
and cross-checking against reality is stopped. In this case, all the objective
events are distorted in the conscious mind and are dogmatically classified
under one of the "working hypotheses"; the influx of new, additional in-
formation virtually ceases (an excellent example is the religious dogmatism).
The strength of "working dogmas' is in their very mutability: they are
constantly adjusted to fit the current level of science. The "aims" of a
civilization can be determined only if complete information on the funda-
mental properties of complex evolving systems is available. Unfortunately,
no such information is available at this stage. The rapid development of
Science gives grounds for hoping that in the near future the theory will be
in a position to advance definition "recommendations" regarding the course
of development of the entire human civilization. Implementation of these
recommendations will be the task of a united, harmonically developing and
"self-regulated'" society.

We are in no position to choose between the two basic alternatives —
expansion into outer space and creation of an artificial autoevolutionary
Lem's world. It is easier to analyze the deficiencies of the various alter-
natives than to propose specific means of their implementation.

251

EXTRATERRESTRIAL CIVILIZATIONS

To conclude our brief review of Lem's book, we would like to emphasize
again the great importance of the methodological approach advanced by Lem
for the solution of a wide circle of problems related to the search for
extraterrestrial civilizations.

The-methods discussed in the previous part of the chapter are applicable
not only to the "general theory of civilizations," i.e., the analysis of the
fundamental properties of complex systems. This technique is also fruitful
in application to "particular" problems. One such problem is the
possibility of "natural" formation of complex systems in the Universe.

A characteristic example of the "cybernetic" approach to the problem of
the origin of life in the Universe is provided by Taube's work /15/.
Proceeding from Lyapunov's functional definition of life /16/, Taube
considers all the "natural" processes and material objects in the Universe
which could provide the raw material for the creation of a living organism.
Various necessary conditions are taken into consideration, such as suf-
ficient abundance of certain elements, the ability of various compounds

to combine into structures and to fulfill certain service functions (transfer
of high- and low-entropy energy, transmission and storage of information).
Taube came to the conclusion that the only material carriers of life under
natural conditions are molecules of hydrogenous compounds which are
spontaneously synthesized in an inanimate environment. The use of
compounds without hydrogen, oxygen, and carbon as the "buiding blocks"
of life forms is ruled out for fundamental reasons by the author. Let us
analyze in some detail the validity of Taube's conclusions. First, Taube
arrived at a precise formulation of the problem from the point of view of the
functional principles of systems. He then investigated a large class of
phenomena which could fulfill the functions of a "living system." Further-
more, he considered (although partially) the exact conditions under which
complex "living" systems may originate in nature.

This approach is free from the "anthropocentric" bias of the studies
concerned with protein life forms. In any case, Taube tries to prove
the "universality" of the protein life form, as one of the very few per-
missible alternatives under the typical conditions prevailing in the
Universe.

Taube's conclusions regarding the possible existence of "life" in the
Universe are thus more objective. Using this approach, we can establish
the possible "external morphology" of systems qualifying for the adjective
of "living" and thus obtain more precise criteria for differentiating
between the "animate" and the "inanimate" in the Universe.

One of the topics considered in connection with the problem of extra-
terrestrial civilizations is the possible impact of an encounter with
intelligent beings from other planets. As a rule, the answer to this
problem is formulated within the framework of "anthropomorphic" concepts.
Extraterrestrial civilizations are considered with regard to their "humane-
ness" (or, conversely, "aggressiveness") /A/. The difference between
civilizations is treated from purely quantitative "orthoevolutionary' aspects.
Note that Stapledon /17/ was the first to consider in detail the problem of
encounter with "differently made" civilizations.

A direct consequence of the "anthropomorphic" approach is the idea
of "interplanetary aid' to be extended by civilizations following such an

encounter.

252

VI, GENERAL TOPICS

In the light of our previous analysis, the encounter with other civiliza-
tions may be regarded as a characteristic "competition" between different
intelligences. More "rationally constructed" systems are characterized
by a higher adaptability, and this fact may have a decisive influence on
other civilizations. The situation is not unlike that discussed on p. 248
in connection with the concept of "autoevolution." Having "become aware"
of its "nonrational" makeup, a civilization will certainly put this infor-
mation to work in order to improve itself. Failure to take any action
because of "unacceptability" of the alternative courses of development of
human society would be tantamount to the acceptance of the theological
thesis concerning the uniqueness of humanity and its "predestiny." An
outcome of encounters with extraterrestrial civilizations would therefore
be acquisition of "purely scientific information' regarding comparative
characteristics of the principle of action of other systems. From the
methodological point of view, any results suppressing the anthropocentric
elements in our scientific thought will be most valuable. Speculations
on the subject of possible "conflicts" in interstellar encounters we leave
to science-fiction writers.

$4. THE PROBLEM OF EXTRATERRESTRIAL
CIVILIZATIONS FROM THE POINT OF VIEW
OF THE GENERAL THEORY OF SYSTEMS

New scientific disciplines falling under the category of cybernetic
methods developed as a generalization of principles which are still being
used by science in the study of the reality around us. The analysis of
phenomena of highly complex structure necessitated a revision of the
basic principles of construction of scientific methods and analytical
techniques. Therefore, the cybernetic approach does not introduce a
"new way of thinking" into science, but the more accurate definition
of the fundamental concepts opens a new and more effective way to tackle
the most entangled problems in natural sciences. The value of correct
methods of scientific research, even if they are confined to the rigidly
"traditional" classical methodology, is in no way reduced by the discovery
of new generalized principles. This idea regarding the continuity of
Scientific methods is best illustrated by the following example. In the
previous part of the review, we criticized a certain conception of the
problem of extraterrestrial civilizations. Our critique, however, did not
weigh some hypothetical "cybernetic conception' against the "classical"
approach, although in the course of the discussion we did mention the need
for a systematic approach to the analysis of the problem. We mainly
questioned the underlying premises of the "energy" hypothesis. On the
other hand, the other particular problems relating to the existence of
civilizations are solved correctly, and the "classical" solutions generally
coincide with those stemming from "cybernetic" principles.

Thus, in his analysis of the distinctive features of the radio waves
from suspected "artificial" sources, Siforov /18/ concentrated on the
statistical structure of radio signals. He writes that "in particular,
if the received signals are narrow-band signals, it is advisable to determine
the two-dimensional probability density distribution of the end of the vector

253

EXTRATERRESTRIAL CIVILIZATIONS

describing the amplitude and the phase of the incoming oscillations in a
plane. The surface describing these two-dimensional distributions provides
an indication regarding the use of feedback in signal generation. It seems
to us that the study of the statistical structure of the incoming signals
will prove useful in deciding whether these signals are
"artificial" or are generated by natural processes not related to the
activity of intelligent beings."

We would like to call the reader's attention to the analogy between
this approach and the method of the "black box" (p. 259), which is one of
the fundamental techniques of modern cybernetics. A consistent application
of this method in passing from restricted problems (the properties of the
radio signal generator as inferred from signal statistics) to more general
topics (artificiality criteria) appears to be quite promising.*

Before we advance further with our analysis, we shall have to introduce
a number of concepts relating to "new" scientific disciplines, such as
cybernetics, information theory, and others.

The term "system" in cybernetics represents an interrelationship of
various elements which are described by sets of significant variables.
Discovery of systems corresponding to this definition is linked up with
the analysis of interrelated phenomena in the Universe. The main
emphasis is placed on the principle of interrelationship, and not on
particular cases of systems represented by certain material constructions.
Examples of systems fitting this definition are the atomic nucleus, the
solid state of an object, language, a game of chess, a conversation between
two friends, etc.

An important point is the possibility of classification of systems.

This classification is generally built according to the degree of complexity
of the system, Moreover, the classification can be based on other
principles also, e.g., systems may be classified according to the
nature of the binding forces, namely deterministic and stochastic systems.

The development of the concept of a system and its properties leads
to the definition of a "machine." A "machine" is a system whose state
changes so that the state variables are interrelated by a certain trans-
formation law. In accordance with the specific character of the system,
we distinguish between deterministic and stochastic machines with simple
and complex laws of transformation of the current parameters. A machine
can be interpreted as a "target-oriented" system, i.e., a system whose
organization is triggered in a sense to fulfilling the tasks that it is entrusted
with /20/. The word "task" is not to be understood as "the goal set up
by another system" (in particular, "man"), but only as the functional
principle of the system.** Machines according to this definition cover a
wide range of phenomena, covering atoms of the individual elements to
planetary systems, cells and tissues of the living organisms, living
organisms themsleves, the "biosphere," and even the biological evolution
as a whole.

^ This methodological approach to the problem of extraterrestrial civilization apparently was first advanced
by Golei /19/.

* How to avoid in cybernetic treatment assigning conscious target or mission orientation to systems is a very
important problem, not only from the viewpoint of the construction of a correct "metalanguage" for the
description of some properties of complex systems. but also for elucidating the objective significance of
such phenomena as "consciousness," "psychology," “purpose of existence" of a complex system (see /21/).

254

VI. GENERAL TOPICS

An extraterrestrial civilization may be treated as a system or a
"machine."

An important characteristic of machines is the character of the dynamic
coupling between the different parts of the system, e.g., the presence of
positive or negative feedback.

S. Lem /1/ considered the remarkable classification of machines
proposed by de Latille. De Latille distinguishes between three principal
groups of machines according to the mode of their operation. The proper-
ties of the representatives of each successive class includes the properties
characteristic of the previous classes. The first group of de Latille's
classification (deterministic systems) includes simple and complex tools
(non-automatic devices) and systems without feedback coupling to the
environment. The second class includes organized regulated and self-
regulated systems with feedback. This wide group of systems covers
mechanical automatic regulators with feedback, programmed machines
and self-programming installations (including man and animals,
regarded as individual representatives). The third group includes
systems which may change their structure and their ''functional principles"
using appropriate input materia). The best example of this group is
provided by biological evolution. De Latille also suggests the existence
of a fourth group of systems, which are additionally endowed with freedom
and ability to choose appropriate components from the environment in order
to "build themselves up." The scientific and technical activity of mankind
as a whole is obviously a system belonging to this group. In Lem's
opinion, de Latille's classification can be extended to cover still another
group of systems, namely those which do not select the input material
for "self-organization'" from the "naturally existing" resources in the
Universe and do not apply the physico-chemical technology to manufacture
the required synthetic materials, but rather create "synthetic" conditions
which are never generated by natural physical processes. We are thus
entering the domain of artificial creation of new forms of existence of
matter, which according to Lem will be one of the attributes of mankind.

Classifications like the one above are of the greatest importance.

They define the position of the system being considered among all the
other systems using certain fundamental features, and thus permit
formulation of the problem of analysis ofthe system ina suitable perspective.
A clear formulation of the problem is especially essential in our search
for extraterrestrial civilizations.

The next important step in the general classification of cybernetic
concepts is the generalization of the concept of system stability. The
principle of homeostasis plays an important role in this respect
/20,21/. According to the homeostasis principle, a target-oriented
system functions in such a way that the values of certain significant
internal variables are maintained between certain limits, despite a variety
of (regular or irregular) external stimuli.

A homeostat according to this definition is a machine with an adequate
regulating mechanism which sustains all the "critical life parameters"
at a certain level. All the phenomena in the animate world are essentially
homeostatic. The concept of homeostasis was actually introduced following
a generalization of the results of biological observations. A discovery of
“homeostatic behavior" in a system is therefore of the greatest importance
for elucidating the exact nature ofthe particular phenomenon. Extraterrestrial

255

EXTRA TERRESTRIAL CIVILIZA TIONS

civilizations apparently also can be regarded as highly complex stochastic
Systems of homeostatic nature.

The fundamental nature of the homeostasis principle is further
Stressed by the fact that this is one of the very few clearly formulated
conceptions which specify the probable "target" or "goal" of the evolution
of complex self-organizing systems. Such properties of living systems
as adaptability, survival, consciousness are on the whole governed by
the principle of homeostasis. These properties therefore can be
considered as a subsystem of the effective regulator incharge of sustaining
the overall homeostasis of the system.

The concept of "intelligence" is of paramount importance, as we have
Seen, in any attempt to define effective artificiality criteria in the search
for extraterrestrial civilizations. We have noted before that the methods
of cybernetics provide a means for the construction of a functional
definition of "intelligence." 'This definition emerges from the theory of
complex self-programming and self-organizing systems, which are no
longer very far beyond the reach of modern science.

A prerequisite of "intelligence" is primarily the ability of a system
to store and process information. In the most general sense, information
can be defined as a measure of ordering, a measure of the decrease
in the uncertainty of the state of the system. In this sense, any machine
has information, since its characteristic law of transformation limits
the variety of other alternatives (states) which are thus unfeasible.
Therefore any machine can be treated as an information processing
machine /22/. This is indeed one of the fundamental principles of cyber -
netics. The modern theory of information deals with quantitative measure-
ments of information and means of optimum information transmission.

We are interested, however, in a slightly different aspect of the theory,
namely what methods of information storage and processing are charac-
teristic of "high-order" systems, i.e., what are the sufficient signs of
"intelligence?"

One of the distinctive features of information transmission is that
information is transmitted in coded form /20/. Information can be stored
in the system in coded form, constituting a "memory bank' of the system.
A suitable example is provided by the storage of genetic information in
biological evolution. Proceeding from the available forms of information
Storage and transmission, we can move on to a more complex concept,
that of the "logic of the system." The logic of a complex highly organized
System is to be understood as its ability to reflect the external processes
of the enivronment* by means of a certain set of internal responses
presentable in a coded form and to apply these sets of states to the
analysis and forecasting of external situations with the purpose of sus-
taining and "improving" the homeostasis of the entire system. The
existence of a special "logic unit" is thus assumed, which operates with a
set of coded symbols ("concepts"). The human brain is clearly one of
these logic units. We are currently in a position to intelligently discuss
the various "forms of logic" characteristic of complex automata, and we

* The very structure of the internal parameters may be interpreted by the logic apparatus of the system as a
manifestation of an "external" situation, i.e., as an object for logical analysis (e.g. the study of human
anatomy by man). This extension of the concept of "environment" is essential to avoid imposing
restrictions on the possibilities of the analytical apparatus of the system.

256

VI. GENERAL TOPICS

are thus probably not far from a reliable classification of the distinctive
features of the various "logics," according to the structure and the functional
principle of the system.*

The "cybernetic" approach enables us to advance a definition of an
"intelligent" system. We can tentatively define an extraterrestrial
civilization as a highly complex stochastic machine of homeostatic
character equipped with the required mechanisms in the form of "logic
units" for information storage and processing, ability to analyze various
situations and to apply the results of this analysis for purposes of directed
evolution, in accordance with certain principles of directed action.

As we have stressed several times, the functional character of
the cybernetic definition is the main feature. The feasibility (at least
theoretically) of the functional definition forces us to advance a lucid
formulation of our aims in the search for extraterrestrial civilizations.

The functional definition of a "civilization system" rules out the
"anthropomorphic" approach.** The class of extraterrestrial civilizations
encompasses not only "anthropomorphic" civilizations, but any other
forms of "intelligent" existence, as long as they possess a sufficiently
varied selection of parameters required for sustaining the programmed
target-oriented activity. There is no more need for the various restrictions
imposed on the search for the possible manifestations of "intelligent
activity" by a certain class of typology, and the ambiguity of the
statements regarding the degree of reliability and single-valuedness
of the interpretation of critical '"difficult-to-explain' phenomena is
automatically eliminated. The same naturally applies to the artificiality
criteria of radio signals. In principle, this presents us not only with an
opportunity to "decode" semantic information, but also to determine the
origin of the signal by gradually refining the methods of structural
analysis of the signal.

* This line of reasoning shows that we will be hard pressed indeed to define a clearcut boundary between
“intelligent” and ‘unintelligent systems, This is further borne out by some findings of modern biology,
which point to the existence of certain elements of "consciousness" and "logic" in various animal
species, This only provides additional proof of the functional character of the very concept of “intelligence,”
which is based on a purely material foundation — the structure and the presence of cerrain mechanisms,
Such concepts as “consciousness,” “emotional response,” etc., are related to certain properties of complex
systems 21/,

The traditional intuitive approach maintained that “consciousness,” “logic,” and “emotions” are the
principal and decisive attributes of an "intelligent" system, Modern science opens new ways for the
interpretation of the strictly “utilitarian” significance of such aspects of civilizations as religion and art.
Art in the light of the theory of complex systems may be interpreted not only as a means for acquiring
additional information, but as a "teaching" or "training" process regulating and controlling the "emotional"
and "aesthetic" properties of the complex highly organized systems /1/. In any case, "tuning" phenomena
in the complex system corresponding to human civilization may and should be interpreted as phenomena
which are objectively connected with the "functional principle” of the system, and their properties should
be elucidated through a study of their functional significance for the system /1/. All of historical
materialsim is based on this point of view.

** In any case, the "anthropomorphism" of the relevant statements is “lowered” to such a level that we can
reason "nonanthropomor phically" concerning the laws which govern the world around us. Human
consciousness typically analyzes the world by means of a certain logic apparatus (which includes, e.g.,
"mathemartization" of the methods of analysis), To speak of "anthropomorphism" at this level is to
maintain that the outside world is arranged "chaotically," without any “causal relations," etc. These
statements clearly contradict the materialistic theory of knowledge.

257

EXTRATERRESTRIAL CIVILIZATIONS

The functional definition of "civilization" suggests some general
principles for the treatment of the "theory of extraterrestrial civilizations."
The problem is reduced from one of "astronomical" importance to a typical
"terrestrial" problem. The advances in the theory of complex systems
regarding the fundamental properties of highly organized forms of existence
should provide a proper foundation for the development of a valid "particular"
theory of highly organized extraterrestrial systems. In this respect, the
importance of theoretical cybernetics in the study of extraterrestrial
civilizations is analogous to the contribution of theoretical physics, say,
to modern astronomy. The theory of extraterrestrial civilizations,
on the other hand, may contribute to the development of cybernetic concepts,
e.g., through analysis of the specific conditions prevailing on various cosmic
objects.

The cybernetic techniques apply to a wide range of effects. In cyberne-
tics, any complex system can be Studied by the "black box" method.

A "black box'' is a mode of a system which is to be studied without any
information being available beforehand about its internal structure. This
System can be simulated by a machine with an "input" and an "output."

The "input" is the complete range of stimuli and interactions to which the
System is exposed, whereas the "output" comprises the various responses
of the system to specific input stimuli. In principle, a real system may
have an infinity of "inputs" and "outputs." The number of "inputs" and
"outputs" in a certain sense is determined by the method and the latitude
of the experiments to which the object is subjected.

Before proceeding with a detailed discussion of the "black box" tech-
nique, we would like to discuss briefly the highly fruitful concept of models
in science. F. Engels, more than a century ago, published a brilliant
analysis of the genesis of scientific knowledge as related to the practical
activity of mankind and called attention to the fact that the unknown or
ungrasped phenomena, which are "things in themselves," are converted
into a fully known "thing for us' once we succeed in reproducing the
corresponding phenomenon artificially. The transformation from a
"thing in itself" into a "thing for us" is a lengthy step-by-step process
in the course of which we investigate one after another the various new
features of the phenomenon, gradually approaching full knowledge of all
the basic properties of an objectively existing "thing in itself." This
reasoning is the basis of the modern scientific concept of a "model."

The methods of modern science (e. g., physics) almost invariably make
use of particular models of the phenomenon for purposes of mathematical
description. All the theories of modern physics are essentially based on
certain physical and mathematical models. If the model fully corresponds
to the original and it covers all the properties of the source object, the
model is said to be isomorphic to the original. In this sense, for example,
we can say with complete certainty that the representatives of one class
of objects are fully isomorphic to one another. Thus any two hydrogen
atoms are fully isomorphic to each other. From the standpoint of cyber-
netics, however, a model is isomorphic to the original object as soon

as it faithfully duplicates all the operations and functions that the original
object performs; there is no need to demand complete similarity of

the model and the object in cybernetics.

258

VI. GENERAL TOPICS

Models of complex systems generally are not fully isomorphic to the
real object. This leads to the concept of homomorphic models. A homo-
morphic model corresponds to the original phenomenon to a degree, on
a certain level, and provides a correct interpretation of a limited range
of properties of the original phenomenon. For instance, an electronic
computer is homomorphic on a certain level to the human brain, since
it performs a number of definite logic operations, although functionally
the computer is basically different from the human brain. It is significant
that were we able to devise a machine capable of performing all the various
operations of the human brain, so that the "potentialities" of the machine
and the brain would be identical, the result would be a cybernetically
isomorphic model of the brain. This again emphasizes the special
importance attached in cybernetics to the functional principles of processes,
rather than to the particular material expression in the form of a "thing"
with all its various external signs and features.

We are now ready to present the principles of the "black box" tech-
nique. The "black box" approach is applicable to highly complex methods
whose structure is inaccessible to direct study.

Applying certain stimuli to the system input (which is equivalent to
the various interactions of the system with the environment and with other
systems, which need not be artificial experimental devices"), we can
study their functional relationship to the output responses of the "black
box." At every stage, a model homomorphic to the actual phenomenon is
created (e.g., in the form of a working hypothesis). The principal aim
is to establish the law of transformation from "input" to "output," i. e.,
the functional principle of the machine. Accumulation of data permits the
construction of homomorphic models on progressively more sophisticated
levels. In this limit, an isomorphic model of the phenomenon will be
obtained. The basic features of this approach lead to an apparently
paradoxical conclusion: in principle, full and exhaustive information can
be obtained about the "black box" without in any way disclosing its actual
physical structure. The missing link, however, emerges directly from
the definition of an isomorphic model. An isomorphic model is functionally
equivalent to the real system. This model is completely interchangeable
with the real object, since it performs all the analogous functions. The
construction of an isomorphic model corresponds to artificial duplication
of the real phenomenon. It was Engels who first suggested basing the
criterion of transformation from a "thing in itself" to a "thing for us" :
on the possibility of artificial duplication!

To apply the "black box" approach to the problem of extraterrestrial
civilizations, we have to discuss the information flow process in obser-
vations using the "black box."

In addition to the above, we should remember that a flow of information
is possible only in certain systems, where the different system components
are linked by communication channels.

The "black box" technique can be represented in the form of a certain
information machine, where the object and the observer constitute a single
system with a feedback loop (Figure 72). It is significant that the information
is transmitted in coded form. Therefore, if the different system compo-
nents use different codes, suitable code-translating units should be provided,
converting the code of one component into a code "understandable" to the

259

EXTRATERRESTRIAL CIVILIZA TIONS

other component. Figure 72 thus shows a block diagram of an information
processing cycle in the "black box" technique. The stimulus applied at the
"black box" input determines the output response in the form of a certain
signal.* The observer "decodes" these signals and interprets them ac-
cordingly. 'The feedback loop is provided by the observer who tests his
hypotheses and conclusions by applying new stimuli to the system input.
The choice of the input stimuli delivered to the "black box" is essentially

a process of translation of the "signal" originating from the observer

into a form "understandable" to the "black box." Each information exchange
cycle provides some additional data to the observer who, having accumulated a
sufficient quantity of information, will create a "black box" of an appro-
priately high level.

Output

Input

Black bos

. : Information (signal)
Signal in “black box in "black box" code
code (input stimulus)

Observer

r----

ro uu Fa

Decoder Cinter-
pretation of
"black box"
signals)

Signal in ob-
server code

"Translator"
(choice of input
and stimulus)

Accumula-
tion of
information

1 ——. —

— me ee ee -æ ee ee —À — — —À — — —

Model
constructicn

Cie ee ee m e o e a

-———T----------

Ecl

FIGURE 72. Observations by the method of a "blaci' box” with an input,
diagramcd in the form of an information machine.

This constitutes a schematic description of a "learning model," illustra-
ting the process of analysis of an unknown effect by the usual scientific
methods.

The "black box" approach is particularly useful for the analysis of
phenomena on distant extraterrestrial objects. The problems of astronomy
essentially reduce to the investigation of inaccessible "black boxes."
Astronomers of all ages seem to have been applying this technique, without
realizing the cybernetic significance of what they were doing.

* The concept of a "signal" is a fairly comple» onc, and it often lends itself to an ambiguous interpretation.
In the most general sense. “signal” is to be understood as a mode of information tranenission. This defi-
nition does not restrict us to a matcrial information carrier, a “material” interaction between the two
parts of the information machine. For example. in an “intelligence machine,” the lach of interaction
is also a kind of signal, since it indicates (carries information about) the abscnce of certain particular
phenomena. (For example, the absence of "cosmic wonders" is a significant picce of informarion in the
theory of extraterrestrial civilizations.)

260

VI. GENERAL TOPICS

In astronomy, however, the "black box" problem is somewhat more
complicated than before. The astronomical objects in a sense are "black
boxes" without "input." Because of the tremendous distances to these
objects in space, we only detect the output signals, which are invariably
in the form of electromagnetic radiation. The astronomer cannot
"experiment" with the object by altering the conditions of the phenomenon.
Thus there is no feedback in the "observer—object'" system (Figure 73).

Output

Black box

Signal from
"black box"

Observer

Computed charac-
teristics of the
model “signal

Comparison of
model “signal”
and “black

box" signal

Final homo-
morphic model

Correction of signal (result of
comparison of computed and
i actual signals)

FIGURE 73. Ohservations by the method of a “black box" without an input,
diagramed in the form of an information machine.

The cyclic flow of information in this machine is confined to the
subsystem "observer." The output signals of the "black box" are compared
with those which would presumably constitute the output of some hypothetical
model of the phenomenon. In actual practice, the entire astronomical
reasoning is based on analogies. The multivalued choice of models may be
limited by analyzing classes of analogous phenomena. The "predictability"
of new aspects and features of the phenomenon plays a very important role
in the process of decision making. However, from the point of view of
our cybernetic concepts, the astronomical research methods essentially
amount to successive "rejection" of inappropriate homomorphic models.
The construction of the isomorphic model, on the other hand, encounters
fundamental difficulties. The lack of the "observer — object' feedback
limits the procedure to the construction of a homomorphic model of a
certain level. As a result, the astrophysicists are forced to consider the
new possibilities of signal detection from extraterrestrial objects and to
improve the existing methods. Particular stress has been placed recently
as a result on the analysis of "fine" effects.

The neglect of the basic fact that no isomorphic models exist in astronomy
often places the astrophysical theories on an excessively speculative basis.
On the one hand, the astrophysicists often have to reluctantly abandon their
traditional conceptions; on the other hand, they are much too keen on
"fashionable" effects and attempt universal interpretation of the various
phenomena from the standpoints of ine particular theories which happen
to be in vogue at the perticular time.

261

EXTRATERRESTRIAL CIVILIZATIONS

Modern cybernetics still has not developed a theory of analysis
of "black boxes" without an input. It is, however, advisable to introduce
some of the methodological ideas of this science into astronomy. It may
yield an additional criterion for testing the reliability of the multitude
of theoretical assumptions regarding our knowledge of the Universe.

Some recent publications show applications of the "quasi-cybernetic"
approach to the analysis of certain astronomical phenomena.

Thus Gudzenko and Chertoprud /22, 23/ tried to investigate solar
activity by a method based on the analysis of the statistical properties of
the "signal," namely the time-dependent parameters of solar activity.
This analysis elucidates the functional principles of mechanisms responsible
for the appearance of solar activity (e.g., whether or not this is a self-
sustained oscillatory system). This approach, in our opinion, is markedly
superior to the traditional observational methods, which primarily search
for particular carriers of activity and only then try to fit it with a plausible
model explaining the functional features.

The cybernetic approach is of special significance in the problem of
extraterrestrial civilizations. In principle, progressively more detailed
studies of the structure of signals from a "no-input black box," based
on the block-diagram of Figure 73, produce homomorphic models of
progressively higher levels. Our understanding of the functional prin-
ciples of the complex system is improved correspondingly. Theoretically,
we can construct a classification typology of functional properties of
progressively increasing complexity; this approach permits assigning
every individual object to a certain class of the classification.

For example, a very extensive class of objects comprises systems
with feedback of all kinds and of all degrees of structural complexity.

The appropriate data can be obtained even now, by a detailed analysis of
the statistical structure of the incoming radio waves, say. The next
stage would be to try and define a narrower subclass of objects displaying
homeostasis. Finally, a group with even more complex functional
properties will be isolated from the homeostatic class. Ultimately,

we can visualize in principle a class of objects which are homomorphic
on a sufficiently high level to the Earth civilization. This is a fantastically
difficult job, and it will not be solved in less than a few years or even
decades. The attractive prospect of this approach, however, is that it
permits formulating in precise and consistent terms the actual purpose
of research. This approach eliminates the uncertainty inherent in the
interpretation of effects within the framework of their classification into
"artificial" and "natural." Each effect will now be regarded as represen-
tative of certain particular features of the generating mechanism. If we
can prove that these "features" define a certain "logic system," we shall
have discovered an extraterrestrial civilization.*

* This method is intrinsically "minimalistic" in its evaluations. Having established that a certain object
can be classified in terms of the distinctive features of its output signal in the lowest class of the typology,
we conclude that this object definitely belongs to that ciass, without, however, ruling out the possibility
of its being part of some higher class of the same typology. This frame of reference is clearly the most
adequate for astronomical] research, because of the impossibility of proving the isomorphism of the result
to the actual phenomenon.

262

VI. GENERAL TOPICS

Finally, the class of "civilizations" in principle may contain systems
with greatly differing morphological features (e.g., theoretically we can
envisage ''non-technological" systems with an entirely different set of
elementary concepts, etc.).

It is naturally irnportant to take into consideration all the additional
information concerning the properties of the proteinic life forms, the
Specific conditions in space in different parts of the Universe, anda
variety of other data which are currently used to a varying degree in
the formulation of the problem of extraterrestrial civilizations (including
the search for "anthropomorphic" civilizations! ).*

In principle, the search for extraterrestrial civilizations should
proceed according to the following methodology.

We have to concentrate on the various aspects relating to the functional
principles and the basic laws of behavior of very complex systems (covering,
in addition to structural analysis, the problems of evolution and forecasting
of the future forms of existence of the system). The aim is to create a
reliable classification of systems according to significant distinctive
features. All these are topics which fall within the competence of
"terrestrial sciences," specifically the theory of complex systems. A
very difficult problem is the determination of the "artificiality criterion,"
i.e., a system of signals which can be identified unambiguously as origi-
nating from a highly organized system ("civilization").** These signals
Should have certain distinctive features which can be applied to differen-
tiate them from other signals, however complex, which originate from
astronomical objects that cannot be regarded as "civilizations." From
the point of view of cybernetics, this amounts to the determination
of the level of organization of a "black box" without input from its output
signal /24/ or the synthesis ofthe originating system from the characteristic
features of the observed signal.

Certain signal sequences (structures) can be theoretically devised
such that the originating system will of necessity have a number of highly
complex features (e.g., a "memory," ability to "recognize patterns,"
create abstractions, etc.). These signal sequences constitute regular
Structures organized in a special manner according to definite set-
theoretical principles, as if to "demonstrate" certain functional
properties of the system enabling it to perform complex operations
/25/. The development of these topics is again not a strictly astronomical
problem.

" There is a possibility that the search for an “anthropomorphic” civilization will culminate in the discovery
of a semantic system of communication signals. This possibility does not detract from the generality of
our conclusions. The above reasoning should not be interpreted to indicate that civilizations of this kind
cannot be discovered, since after all the subclass of "anthropomorphic" systems is part of the corresponding
cybemetic typology. Ifa "man-like" intelligence were to be discovered in the form predicted by the
“anthropomorphic -energetic" hypothesis, this would justify the assuinption of the universal applicability
of the "technological" and cnergetic mode of development, but only on grounds of correspondence to some
deeper underlying principles of the growth of complex systems.

** [lere we naturally ignore the question of the existence of “anthropomorphic communication" with
semantically decodable information. This particular case provides an unquestionable “artificiality
criterion.” However, "semantic communication” requires sufficiently long “Messages” adequate for
successive decoding of the individual symbols, so that we again return to the problem of "call signals,”
i.e., sufficiently brief endings which are highly effective for detection purposes (sce Chapter ID.

263

EXTRA TERRESTRIAL CIVILIZA TIONS

Depending on the success of the above measures, we will have to
continue with a consistent study of the "functional principles" of astronomical
objects, after clearly formulating exactly what properties we are interested
in discovering. This is evidently an astronomical problem. It also
includes a generalizing part: elucidating the effect of various cosmic
conditions on changes in functional principles of complex systems.

Other branches of science, e.g., radio astronomy, certainly can make
their contribution to the analysis of individual problems. The study of the
various "noises" interfering with proper signal detection and distorting
the signals constitutes a separate part of the comprehensive overall in-
vestigation.

All the available scientific data lead to the conclusion that a precise
methodology can be devised for the solution of the problem of extra-
terrestrial civilizations on the current level. The great complexity of the
problem stems from the fact that it is inseparably linked with even more
fundamental problems. Therefore, only further advances in the methods
of cybernetic analysis, the general theory of systems, biology, and
other disciplines will enable significant progress to be made toward the
Solution of the problem of extraterrestrial civilizations. We cannot rely
on a "lucky chance" that will enable us to "guess" the answers to the main
questions, which are not even always clearly formulated. The same also
applies to experiments aimed at detection of astronomical signals
bearing signs of "artificial" origin. Here the primary problem is clear
and precise formulation of the "artificiality criteria" and detailed analysis
of a very extensive class of astronomical phenomena. On the whole, this
is not a fundamentally new problem. S. E. Khaikin notes that the problem
of systematic search for radio signals of artificial origin on the whole
coincides with the fundamental problem of radio astronomy: accumulation
of information about the cosmic radio sources /26/. The differentiation
will become possible only after the application of "artificiality criteria"
to particular classes of objects.

The problem of the "general theory of civilizations" will clearly be
one of the major subjects of contemporary and future science. Any
progress toward the solution of this problem is predicated on the general
advancement of science. There is no doubt that this field of research
will eventually occupy a prominent position among the other scientific

disciplines.

Bibliography

Lem,S. Summa technologiae. — Krakow, Wyd. Lit. 1964.

2. Shklovskii,I.S. Vselennaya, zhizh', razum (Life and Intelligence
in the Universe), 2nd Edition. — "Nauka." 1965.

3. Kardashev,N.S.— Astron. Zhurnal, Vol.41:282. 1964.

4. Vnezemnye tsivilizatsii (Extraterrestrial Civilizations). Proceedings
of a Conference, Byurakan, 20—30 May 1964. — Izd. AN Arm.
SSR. 1965.*

5. Cameron, A. (Editor). Interstellar Communication.— New York.
Benjamin. 1963.

6. Lilley, S.—Sociology of Science, N.Y. 1962.

* [Sec footnote on p. 11.1

264

25.

26.

VI. GENERAL TOPICS

Price,D.— Discovery, Vol.6:240. 1956.

Dobrov,G.M. Nauka o nauke (The Science of Science). — Kiev,
"Naukova Dumka," 1966,

Simon, R.— Astronaut and Aeronaut, Vol.3:59. 1965.

Freudental,H. Lincos, Amsterdam. 1960.

Gladkii, A.V.— In: /4/:145.

Baumshtein, A.I.— Priroda, No.12. 1961.

Breisuell,R. In: /5/:271.

Hoyle,F. Of Men and Galaxies. Univ. of Washington Press. 1966.

Taube,M. Hydrogen the Carrier of Life. — Nucl. Energ. Inform.
Center, Warsaw. 1965.

Lyapunov,A.A.—In: "Kibernetika, myshlenie, zhizn'," p.127.
"Mysl'." 1964.

Stapledon,O. Last and First Man.— London, Peng. B. 1939.

Siforov, V.I.—In: /4/:121.

Golei. see /5/.

Ashby, W.R. An Introduction to Cybernetics. — Chapman and
Hall. 1956.

Bir,St. Kibernetika i upravlenie proizvodstvom (Cybernetics and
Industrial Control).— Fizmatgiz. 1963.

Gudzenko,L.I. and V. E. ChRertoprud.-— Astron. Zhurnal,
Vol.41:597. 1964.

Gudzenko,L.I. and V. E. Chertoprud.-— Astron. Zhurnal,
Vol.43:113, 1966.

Panovkin, B. N. — Doklad na Sessii NTORiE im. A. S. Popova, Moskva,
May 1967.

Panovkin, B. N. - Doklad na 1-oi Konferentsii po kosmicheskoi
radiosvyazi. Moskva. 1968.

Khaikin, S. E.— In: /4/:83.

265

. . oe e POSTAGE AND FEES PAID
National Aeronautics and Space Administration NATIONAL AERONAUTICS: AND

WASHINGTON, D. C. 20546 SPACE ADMINISTRATION

OFFICIAL BUSINESS

70364 00903
ACCRATORY. /WLOL/

C3 021
MEXICO | 57117?

4
APR FCRC

KEATLANE AF GS, NE

ATT E. Loo BOWMAN, CHIEF, TFCH. LIBRARY

Printed in Jerusalem, Israel TT 70-5008]
NASA TT F-63

Internet Archive Audio

Featured

Top

Images

Featured

Top

Software

Featured

Top

Texts

Featured

Top

Video

Featured

Top

Mobile Apps

Browser Extensions

Archive-It Subscription

Save Page Now

Full text of "Extraterrestrial Civilizations Problems Of Interstellar Communication"

See other formats